Hype Cycle for Artificial Intelligence, 2025

11 June 2025 - ID G00828523 - 123 min read

By: Haritha Khandabattu, Birgi Tamersoy
Initiatives:Artificial Intelligence; Analytics and Artificial Intelligence; Build, Engineer and
Implement AI Initiatives; Select and Orchestrate AI Platforms and Technologies

AI investment remains strong, but focus is shifting from GenAI

hype to foundational innovations like AI-ready data, AI agents, AI
engineering and ModelOps. This research helps leaders prioritize
high-impact, emerging AI techniques while navigating regulatory
complexity and operational scaling.

Analysis
What You Need to Know
As AI adoption matures, the focus is shifting from experimentation to scale. Generative AI
(GenAI) is now at the Trough of Disillusionment, signaling a maturing understanding of its
potential and limits. Enterprises are directing investment toward AI-enabling capabilities —
from high-quality, contextualized data to responsible AI governance — to ensure
consistent, scalable delivery. As concerns around AI safety, sovereignty and ethics
intensify, responsible AI is becoming not just a compliance requirement, but a design
principle and a differentiator.

At the same time, newer innovations like multimodal AI, AI trust risk and security
management (TRiSM) adoption, and AI agents (LLM-based) dominate the Peak of Inflated
Expectations. Composite AI has retained its relevance, serving as a foundational strategy
for combining diverse AI techniques. Innovations in simulation, embodied AI and world
models continue to gain traction, offering powerful capabilities that are setting the stage
for future widespread adoption.

Organizations must bring AI closer to the point of decision, using fit-for-purpose AI

solutions tailored to context and workload. Success will depend on tightly business-
aligned pilots, proactive infrastructure benchmarking, and coordination between AI and
business teams to optimize value delivery across environments.

Gartner, Inc. | G00828523 Page 1 of 108

This research note is restricted to the personal use of [email protected].

 The Hype Cycle
The two biggest movers in this year’s Hype Cycle, AI-ready data and AI agents, are
experiencing heightened interest. This trend, accompanied by ambitious projections and
speculative promises, places them at the Peak of Inflated Expectations. As businesses
invest in AI-ready data strategies and deploy AI agents, they must navigate the
complexities of implementation and integration and the risks associated with deployment.
It is essential to manage expectations realistically and plan strategically to get the full
potential of these innovations, while also preparing for the inevitable challenges that
follow.

AI governance platforms and FinOps for AI are two new entries on the Hype Cycle that
indicate a focus on establishing robust frameworks for managing AI’s risks and
expanding its role in enterprises. Their arrival marks a shift toward not only developing
cutting-edge AI technologies, but also implementing structured oversight and financial
efficiency measures to support sustainable growth.

The entry of AI-native software engineering into the Hype Cycle highlights how rapid
changes in the field are profoundly reshaping roles and tasks. Similar to other evolving
roles, AI-native software engineers are redefining their way of working — orchestrating
agentic tools and workflows to design, build and scale intelligent systems with greater
autonomy and creative intent.

Advancements like model distillation, already present in later stages of the Hype Cycle,
emphasize the maturing of capabilities that enhance efficiency and enable more versatile
and comprehensive AI applications.

Meanwhile, innovations such as quantum AI, first-principles AI and causal AI have

experienced slower progression compared to last year’s Hype Cycle. While these areas
hold great promise, their measured pace suggests that they face complex technical
challenges and may require significant breakthroughs before they can be widely adopted
in successful AI solutions.

Last but not least, computer vision has graduated from the Artificial Intelligence Hype
Cycle this year; it has become an established technology with widespread integration in
real-world applications.

Gartner, Inc. | G00828523 Page 2 of 108

This research note is restricted to the personal use of [email protected].

 Figure 1: Hype Cycle for Artificial Intelligence, 2025

The Priority Matrix

This year’s Priority Matrix reflects an AI landscape dominated by innovations of high or
transformational benefit. Compared with 2024, there is a sharper emphasis on operational
scalability and real-time intelligence, with a gradual pivot from GenAI as a central focus
toward the foundational enablers that support sustainable AI delivery.

Within the next two years, edge AI, composite AI and responsible AI are anticipated to
achieve mainstream adoption. Edge AI will enable more efficient data processing by
bringing computation closer to data sources and further increase AI’s potential
applicability. Composite AI will enhance the robustness of AI systems by integrating
multiple AI techniques. Responsible AI will ensure that these advancements are
implemented ethically and transparently, paving the way for trust and accountability in AI
systems.

Gartner, Inc. | G00828523 Page 3 of 108

This research note is restricted to the personal use of [email protected].

 In the two- to five-year time frame, several key AI technologies are expected to reach the
Plateau of Productivity, driving significant advancements across industries. These include
foundational technologies like AI engineering and ModelOps, which streamline the
integration and management of AI systems, and emerging techniques such as generative
AI and multimodal AI, which enhance content creation and data interpretation.
Additionally, AI governance platforms and AI TRiSM are set to play crucial roles in
ensuring ethical and secure AI deployment. Together, these developments will enable more
robust, innovative and responsible AI applications, transforming how businesses and
organizations operate.

Over the next five to 10 years, AI-ready data and AI simulation are expected to significantly
enhance how AI is deployed across industries. AI-ready data will ensure that datasets are
optimized for AI applications, enhancing accuracy and efficiency. Meanwhile, AI
simulation will improve predictive capabilities and scenario testing, and FinOps for AI will
streamline the financial management of AI investments, collectively driving more effective
AI implementations.

Longer-term innovations, such as artificial general intelligence, hold disruptive potential

but face scientific, ethical and geopolitical hurdles. Sustained research and coordinated
ecosystem growth will be critical to their evolution.

AI leaders must balance ambitious exploration with operational discipline — advancing

transformational capabilities while ensuring near-term investments are grounded in
scalable, fit-for-purpose AI solutions.

Gartner, Inc. | G00828523 Page 4 of 108

This research note is restricted to the personal use of [email protected].

 Table 1: Priority Matrix for Artificial Intelligence, 2025
(Enlarged table in Appendix)

Off the Hype Cycle

Gartner, Inc. | G00828523 Page 5 of 108

This research note is restricted to the personal use of [email protected].

 On the Rise
AI-Native Software Engineering
Analysis By: Manjunath Bhat, Mark Driver

Benefit Rating: Transformational

Market Penetration: Less than 1% of target audience

Maturity: Embryonic

Definition:

AI-native software engineering is an emerging set of practices and principles that are
optimized for using AI-based tools to develop and deliver software applications. This
entails AI autonomously or semiautonomously performing a significant percentage of the
tasks across the software development life cycle (SDLC). For example, developers use AI
agents that proactively recommend and execute actions by sensing inputs from
development environments with the goal of automating end-to-end workflows.

Why This Is Important

Today, developers use AI-based tools such as AI code assistants and AI testing tools that
are limited to coding and testing activities. However, Gartner predicts a future where AI will
be integral and native to most software engineering tasks, changing both the nature of
developers’ work and their way of working. AI-native software engineering practices are
important because they enable developers to focus on meaningful tasks that require
critical thinking, human ingenuity and empathy.

Business Impact

AI-native software engineering has the potential to deliver numerous benefits for
developers, product teams and business stakeholders alike. AI-based tools will help
developers go beyond handling drudgery. These tools will serve as ideation partners that
boost human creativity and support product teams in generating new ideas. Product
teams can use AI-enabled analysis to make data-driven product roadmap decisions that
either validate or refute subjective human decisions. Then, teams can expedite creating
prototypes to speed up feasibility studies and chart an informed path forward.

Drivers

Gartner, Inc. | G00828523 Page 6 of 108

This research note is restricted to the personal use of [email protected].

 ■ Achieving step change in productivity and performance: Software engineering has
become AI-augmented with the emergence of AI-based developer tools, but the
overall method and SDLC haven’t yet transformed. AI augmentation has been
applied to many tasks within the traditional SDLC to provide incremental
improvements in lead time, cycle time and quality. This is driving software
engineering leaders to explore ways to achieve a step change in productivity and
performance.

■ Boosting human creativity: Teams can leverage multimodal capabilities in AI-based

development tools to boost human creativity. Instead of staring at a blank canvas, a
product manager or UX designer can use AI design tools to convert text prompts,
screenshots or paper sketches to images and visual prototypes. They can then
iterate on the design and, once the design is finalized, UX designers can convert
those designs to HTML and cascading style sheets (CSS).

■ Translating intent into action: Early previews of AI-native application builders such
as Bolt from StackBlitz, Firebase Studio from Google, V0 from Vercel and Lovable
demonstrate the ability to translate user intent into actions. Users can describe
requirements from an end-user perspective rather than technical specifications. This
signals a fundamental shift in how we express intent to computers — going from
precise instructions to describing the desired state in natural language. Andrej
Karpathy, the cofounder of OpenAI and founder of Eureka Labs, coined the term
“vibe coding” to describe this new programming paradigm where you “forget the
code even exists.” Note that vibe coding as defined is currently limited to
experimental prototypes and not suited to take applications to production.

■ Generating positive developer sentiment: Developers, in general, are keen on

experimenting with new tools. A recent Gartner survey of 5,112 software
development team members from 51 organizations shows that 54% of software
development team members consider having freedom to experiment and innovate
as one of the most important aspects of the developer experience. Positive developer
attitudes toward AI is driving the adoption of AI-native engineering practices within
organizations.

Obstacles

Gartner, Inc. | G00828523 Page 7 of 108

This research note is restricted to the personal use of [email protected].

 ■ Overly trusting AI outputs: AI-native approaches create a new burden on developers
and knowledge workers in general. Developers increasingly offload tasks to AI-based
tools, which carry inherent risks of nondeterminism and hallucinations. Therefore,
blindly trusting AI outputs without verification and explainability can potentially pose
serious business risks, including reputational damage. GitClear’s 2025 AI Copilot
Code Quality Report shows four times more code cloning and a spike in the
prevalence of duplicate code blocks between 2023 and 2024 compared to previous
years.

■ Increased security risk: AI tools expand the threat surface to include the chain of all
events and interactions they initiate and participate in — including those that are
invisible to human or system operators. This expanded threat surface increases the
potential for unforeseen vulnerabilities and security breaches. Developers must
account for agentic workflows as part of assessing and mitigating software supply
chain risks.

■ Compounded risk of hallucinations in multiagentic workflows: The risk of

hallucinations compounds in multiagentic workflows, where AI-generated context is
passed from one AI agent to another. Model overreach is also an issue with agentic
systems — where the model does more than asked because in the training data it
often goes further than just the bit you need. For example, you may ask for it to write
a script for “search” and it creates one that does “search and replace.”

User Recommendations

■ Rethink developer workflows by taking advantage of emerging AI-based tools and

technologies to improve productivity and enhance creative upstream work. Examples
of upstream use cases include product discovery, product design, user sentiment
analysis and feature prioritization.

■ Unlock greater gains in productivity and minimize hallucinations by sharing context

between AI-enabled developer tools (potentially AI agents) as they transition work
from one tool to another.

■ Prioritize low-risk and high-value use cases for agent-based tools by assessing their
ability to automate repetitive work and keep humans in the loop for oversight,
verification and explainability. To sustain high quality, human developers must
review code, use test harnesses and establish security and compliance guardrails
throughout the software development life cycle (SDLC).

Gartner, Inc. | G00828523 Page 8 of 108

This research note is restricted to the personal use of [email protected].

 ■ Explore ways to benefit from autonomous improvement loops by segmenting tasks
based on business criticality, risk threshold and task complexity. Software
engineering leaders will need to look for opportunities where these autonomous
loops deliver business value without increasing risks.

Sample Vendors

Anysphere; CodeStory; GitHub; Google; Lovable; Replit; StackBlitz; Vercel; Windsurf; Zed
Industries

Gartner Recommended Reading

Innovation Insight for AI-Native Software Engineering

Why AI Will Not Replace Software Engineers, and What That Means for the C-Suite

How to Upskill Software Engineering Teams in the Age of AI

Magic Quadrant for AI Code Assistants

Emerging Tech: AI Developer Tools Must Span SDLC Phases to Deliver Value

Quantum AI
Analysis By: Chirag Dekate, Soyeb Barot

Benefit Rating: Low

Market Penetration: Less than 1% of target audience

Maturity: Embryonic

Definition:

Quantum AI is an embryonic field of research emerging at the intersection of quantum

technologies and AI. Quantum AI aims to exploit unique properties of quantum mechanics
to develop new and more powerful AI algorithms that deliver better than classical
performance, potentially resulting in new types of AI algorithms designed to run on
quantum systems.

Gartner, Inc. | G00828523 Page 9 of 108

This research note is restricted to the personal use of [email protected].

 Why This Is Important

Quantum AI is an area of active research. Once commercialized, quantum AI could

potentially help in:

■ Enabling organizations to use quantum systems to address advanced AI analytics

faster while using a fraction of the resources used in conventional AI
supercomputing.

■ Developing new AI algorithms that exploit quantum mechanics to deliver capabilities

beyond ones that can be executed on classical systems.

■ Unlocking disruptive applications that include drug discovery, energy industry and
logistics.

Business Impact

While the business impact of the embryonic quantum AI field today is low, when validated
techniques mature, quantum AI will enable competitive advantage across industries; for
instance:

■ Life sciences: Transform drug discovery by shortening timelines, lowering costs and
improving outcomes.

■ Finance: Optimize portfolios, minimize risk and improve fraud detection systems.

■ Material science: Revolutionize energy transportation, manufacturing and create

new revenue streams by discovering new materials.

Gartner, Inc. | G00828523 Page 10 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ Progress is steady in scaling quantum systems and improving error correction

schemes.

■ Hype around quantum technologies is driving more businesses and researchers to

explore the intersection of quantum and AI.

■ The accelerated pace of innovation in quantum systems (including a larger volume

of higher quality qubits, and greater stability and reliability of quantum systems) is
driving greater interest in applicability in areas, including quantum AI.

■ Access to quantum computing as a service is lowering the barrier to entry,

encouraging greater collaboration among researchers and enabling exploration of
new algorithms and techniques.

■ Governments and enterprises globally are increasing funding for quantum (and
quantum AI) research, resulting in accelerated innovation.

■ The halo effect of increased hype around GenAI is driving new focus on alternative
research techniques, including quantum AI, that could potentially deliver new
disruptive results.

■ Universities and training programs are developing programs and curricula to develop
a quantum-ready workforce.

Gartner, Inc. | G00828523 Page 11 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Hardware limitations: Current quantum systems, while getting stabler, are still error-
prone and inherently noisy, limiting their utility and impact on practical quantum AI.

■ Algorithm limitations: While several quantum AI algorithms have been proposed,

very few have been vetted and proven, and they are nowhere close to being
enterprise-ready.

■ Cost: Despite their limited utility and widespread accessibility, rapidly evolving noisy
intermediate-scale quantum (NISQ) systems are relatively expensive, which could
inhibit research and development efforts needed to devise quantum AI algorithms.

■ Scalability of systems: Scaling quantum systems to the level necessary for

enterprise-ready quantum AI continues to be a major technical hurdle.

■ Compute paradigms: Integrating traditional data and analytics pipelines with

quantum is inherently challenging because quantum systems operate on a
fundamentally different paradigm both from a data representation perspective and
from a compute (non-von Neumann model) perspective.

User Recommendations

■ Prioritize investments in AI and GenAI over any quantum AI investments. Quantum AI

is too nascent to warrant focused investments and unlikely to yield material gains in
the next two to three years.

■ Partner with local universities by sponsoring academic research as a means of

derisking your quantum AI investments and create a university-to-industry talent
pipeline.

■ Create a quantum AI opportunity radar that enables you to track progress of

underlying technologies and quantum AI algorithms, enabling you to maximize value
creation as the embryonic field of quantum technologies evolves.

■ Diversify quantum use cases beyond a narrow AI context into other domains
including materials simulations, search, optimization and other emerging
algorithmic domains.

Sample Vendors

Amazon Web Services; Google; IBM; IonQ; Microsoft; Multiverse Computing; Pasqal;
SandboxAQ

Gartner, Inc. | G00828523 Page 12 of 108

This research note is restricted to the personal use of [email protected].

 Gartner Recommended Reading

Leaders’ Guide to Quantum Computing

Start Your Quantum Strategy Now in Response to Recent Quantum Processor Innovations

Executive Briefing on Emerging Technology: Quantum Computing

First-Principles AI
Analysis By: Erick Brethenoux, Svetlana Sicular

Benefit Rating: Transformational

Market Penetration: 5% to 20% of target audience

Maturity: Adolescent

Definition:

First-principles AI (FPAI; aka physics-informed AI) incorporates physical and analog

principles, governing laws and domain knowledge into AI systems. In contrast, purely
digital AI models do not necessarily obey the fundamental governing laws of physical
systems and first principles — nor generalize well to scenarios on which they have not
been trained. FPAI extends AI engineering to complex systems engineering and model-
based systems, such as agent-based systems.

Why This Is Important

As AI expands in engineering and scientific use cases, it needs a stronger ability to model
problems and better represent their context, but digital-only AI solutions cannot generalize
well enough beyond training, which limits their adaptability. In contrast, FPAI instills a
more reliable representation of the context and the physical reality, yielding more robust
and adaptive systems. This leads to reduced training time, improved data efficiency, better
generalization and greater physical consistency.

Gartner, Inc. | G00828523 Page 13 of 108

This research note is restricted to the personal use of [email protected].

 Business Impact

Physically consistent and scientifically sound AI models can significantly improve

applicability, especially in engineering use cases. FPAI helps train models with fewer data
points and accelerates the training process. As a result, models converge faster to optimal
solutions. FPAI improves the generalizability of models to make reliable predictions for
unseen scenarios, including applicability to nonstationary systems, and enhances
transparency and interpretability, boosting trustworthiness.

Drivers

■ More flexible representation of system context and conditions: FPAI approaches

instill greater context flexibility, allowing developers to build more adaptive systems.
Traditional business modeling approaches have been brittle. This is because the
digital building blocks composing solutions cannot generalize well enough beyond
their initial training data, therefore limiting those solutions’ adaptability.

■ Additional physical knowledge representations: As an example, FPAI approaches

provide physics equations (e.g., partial differential equations) to guide or bound AI
models. AI techniques, particularly in the machine learning (ML) family, have severe
limitations — especially for causality and dependency analysis, admissible values,
context flexibility and memory retention mechanisms. Asset-centric industries have
already started leveraging FPAI in physical prototyping, predictive maintenance or
composite materials analysis, for example.

■ Modeling challenges: Complex systems like climate models, large-scale digital twins
and complex health science problems are particularly difficult to model. Composite
AI approaches provide more concrete answers and manageable solutions to these
problems, but their engineering remains a significant challenge.

■ Simplified and enriched AI approaches: First-principles knowledge defines problem

and solution boundaries, reducing the scope of ML’s traditional brute-force
approach. For example, first-principles-based semantics can reveal deepfakes.

■ Need for more robust and adaptable business simulation systems: With a better
range of context modelization and more accurate knowledge representation
techniques, FPAI simulations will be more reliable and account for a wider range of
possible scenarios — all better-anchored in reality.

■ The advent of systems based on AI agents: The capability of these systems to

promote efficient combination of digital and analog models will, in turn, promote the
use of FPAI.

Gartner, Inc. | G00828523 Page 14 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Development of systematic tests and standardized evaluations for these models

across benchmark datasets and problems could slow the adoption of FPAI
capabilities.

■ Scaling of the training, testing and deployment of complex FPAI models on large
datasets in an efficient manner will be a computational challenge.

■ Collaboration across many diverse communities of physicists, mathematicians,

computer scientists, statisticians, AI experts and domain scientists will pose a
resource challenge.

■ Brute-force approaches are prevalent in AI and easy to implement for data scientists,
while first principles require additional fundamental knowledge of a subject that
calls for a multidisciplinary team.

■ Developers’ difficulty to scope first-principle methods, without access to engineers or

subject matter experts, can also prove a major obstacle for the introduction of these
methods.

User Recommendations

■ Set realistic development objectives by identifying errors that cannot be reduced and
discrepancies that cannot be addressed, including data quality.

■ Encourage reproducible and verifiable models, starting with small-scoped problems.

Complex systems, in the scientific sense of the term, are generally good candidates
for this approach.

■ Enforce standards for testing accuracy and physical consistency for physics and
first-principles-based models of the relevant domain, while characterizing sources of
uncertainty.

■ Promote model-consistent training for FPAI models and train models with data
characteristics representative of the application, such as noise, sparsity and
incompleteness.

■ Quantify generalizability about how performance degrades with degree of

extrapolation to unseen initial and boundary conditions and scenarios.

■ Ensure relevant roles and education in a multidisciplinary AI team with domain

expertise, so the team can develop effective and verifiable solutions.

Gartner, Inc. | G00828523 Page 15 of 108

This research note is restricted to the personal use of [email protected].

 Sample Vendors

Abzu; IntelliSense.io; MathWorks; NNAISENSE; NVIDIA; VERSES

Gartner Recommended Reading

Innovation Insight: AI Simulation

Go Beyond Machine Learning and Leverage Other AI Approaches

Predicts 2023: Simulation Combined With Advanced AI Techniques Will Drive Future AI
Investments

AI Governance Platforms
Analysis By: Lauren Kornutick, Sumit Agarwal, Avivah Litan

Benefit Rating: High

Market Penetration: 1% to 5% of target audience

Maturity: Emerging

Definition:

An AI governance platform provides a central view of AI applications and agents and each
unique use case in the enterprise. It measures performance against agreed-to frameworks
such as policy, regulations and industry standards. In addition to a central repository of
use cases, it consolidates risk management activities, automates workflow approval for
new uses and applies the appropriate level of technical enforcement required in a single-
continuous tool for elements critical to AI.

Why This Is Important

AI is spreading rapidly. AI leaders are increasingly tasked with oversight of AI despite

diverse subject matter expertise required to safely deploy. Oversight also ensures
acceptable use and value based on predetermined criteria. AI governance platforms are
emerging as a distinct market that combines traditional governance, risk and compliance
product capabilities with automating and enforcing essential governance rules to verify
that AI is safe, valuable and performing as intended.

Gartner, Inc. | G00828523 Page 16 of 108

This research note is restricted to the personal use of [email protected].

 Business Impact

We have observed that AI governance has become more centralized, although

organizations may take different approaches. AI governance platforms are emerging to
meet this need and address the governance, risk management and compliance
requirements for deploying AI. AI governance platforms can significantly improve
operational efficiency, engage cross-functional stakeholders effectively and enforce
policies at runtime, all of which contribute to the overall success of AI in the enterprise at
scale.

Gartner, Inc. | G00828523 Page 17 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ The increasingly pervasive nature of AI and a need for transparency and

accountability highlight the need for strong AI governance.

■ Global regulations that directly or adjacently target issues associated with AI, such
as bias, security, safety and ethical concerns, and a lack of transparency and
accountability, increase the complexity of compliance. A consolidated view is
essential for implementing technical controls and oversight through approvals,
validations, audits and observability to allow the enterprise to operate at scale.

■ Organizations create AI governance principles and policies that lack monitoring and
enforcement, which apply to use cases of AI in production.

■ To effectively oversee AI, and deploy at scale, AI governance will require technology
to implement AI governance controls to enact and enforce new enterprise AI
requirements.

■ Successful AI governance requires cross-functional subject matter expertise,

including legal, compliance, risk, cybersecurity, IT and data analytics, to establish
common goals, taxonomies and frameworks, which is difficult to achieve because of
competing priorities. This reduces the effectiveness of governance and increases
risk.

■ Organizations already, and will continue to, do security assessments at the product
level — that is insufficient. AI assessments must be conducted for each AI-enabled
feature and use case.

■ Stakeholders are increasingly demanding trust — and organizations leverage high-

level responsible AI principles and AI acceptable use policies to drive confidence in
an organization’s products, services and the overall brand.

■ Enterprise AI implementations are expanding, incorporating various AI models such

as machine learning, computer vision, large language models and others, and driving
new risks. Enterprises require a platform that provides a comprehensive view of the
enterprise AI portfolio and record of decisions and accountability, its risks, risk
mitigations and reliable performance.

Gartner, Inc. | G00828523 Page 18 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ There is confusion over who should take ownership of AI governance (for example,
legal, cybersecurity, data and analytics, IT.)

■ AI governance teams will overlap with some, but not all, operational policies already
monitored and enforced by data and analytics and IT governance, MLOps and
cybersecurity teams. This necessitates clear perimeters within the organization while
making space for this new function.

■ The diverse nature of AI use cases and priorities requires organizations to decide if a
single common platform of AI governance, or a combination of tools, is most
appropriate for their needs.

■ Vendors are latching onto “AI governance” as a go-to-market strategy due to low
awareness of capabilities. This leaves buyers confused and with products that
overpromise and underdeliver.

■ Global regulations and standards vary wildly — and even though controls needed for
compliance are similar, desired outcomes may be in opposition.

■ Rapid progress makes some features obsolete and calls for new features that lag
the progress.

Gartner, Inc. | G00828523 Page 19 of 108

This research note is restricted to the personal use of [email protected].

 User Recommendations

■ Define acceptable use policies at a level granular enough to enforce for AI use cases.
Acceptable use should be aligned to broader enterprise-responsible AI principles.

■ Establish a review and approval workflow process for low-touch or high-touch

reviews and attestation for new use cases.

■ Establish an enterprise risk tolerance for deployment of AI within the enterprise that
calculates both risk and value in a single risk score.

■ Use the risk score to triage who will need to review and approve the AI and how the
AI will be monitored, including technical controls.

■ Evaluate the AIs performance against risk tolerance and deploy technical controls
for AI. This should be done at some level for all use cases.

■ Determine which governance processes need to translate to technology to effectively

scale AI and to define your requirements for an AI governance platform.

■ Check platform vendor references from your industry. Ensure that the references use
the functionality you require too.

Sample Vendors

Collibra; Cranium; Credo AI; Holistic.AI; IBM; Monitaur; OneTrust

Gartner Recommended Reading

3 Key Steps to Build a Scalable AI Governance Framework

Market Guide for AI Trust, Risk and Security Management

Causal AI
Analysis By: Pieter den Hamer, Leinar Ramos, Ben Yan

Benefit Rating: High

Market Penetration: 1% to 5% of target audience

Maturity: Emerging

Gartner, Inc. | G00828523 Page 20 of 108

This research note is restricted to the personal use of [email protected].

 Definition:

Causal AI identifies and utilizes cause-and-effect relationships to go beyond correlation-

based predictive or generative models and toward AI systems that can prescribe actions
more effectively and act more autonomously. It includes different techniques, such as
causal graphs and simulation, that help uncover causal relationships to improve decision
making.

Why This Is Important

AI’s ultimate value comes from making better decisions and taking effective actions.
However, the current correlation-based generative and predictive approaches have their
limitations. Not only do they offer very limited transparency, generating or predicting an
outcome is not the same as understanding what causes it and how to improve it. Causal
AI is crucial when systems need to be more transparent and reliable in identifying and
prescribing actions to achieve the right business outcomes.

Business Impact

Causal AI leads to:

■ More reliable augmentation and autonomy in decision intelligence or AI-empowered

decision making.

■ More robustness and adaptabilty by leveraging causal relationships that remain

valid in changing environments.

■ The ability to extract causal knowledge, also known as causal discovery with various
AI techniques, sometimes combined with simulations, reduces time and costs of
real-world experiments (for example, A/B tests), although validation is still required.

Gartner, Inc. | G00828523 Page 21 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ In the context of decision intelligence, analytics demand is shifting from predictive to

more prescriptive capabilities. A causal understanding of how to affect predicted
outcomes and optimize decision making is increasingly important.

■ AI — in particular agentic AI — systems increasingly need to act autonomously,

particularly for time-sensitive and complex use cases where human involvement is
not feasible. This will only be possible by AI understanding what impact actions will
have and how to make effective interventions.

■ Limited data availability for certain use cases requires more data-efficient
techniques like causal AI, possibly combined with synthetic data. Causal AI
leverages human domain knowledge of cause-and-effect relationships to bootstrap
AI models in small-data situations.

■ Growing complexity and dynamics of business require more robust AI techniques.

Correlation-based AI models, trained with historical data, are brittle and lose
accuracy when faced with gradual, let alone disruptive, changes. Causal structure
changes much more slowly than statistical correlations, making causal AI more
robust and adaptable in fast-changing environments.

■ The need for greater AI trust and explainability is driving interest in models that are
more intuitive to humans. Causal AI techniques, such as causal graphs, make it
possible to be explicit about causes and explain models in terms that humans
understand.

■ Generative AI (GenAI) can accelerate causal AI implementation. GenAI is emerging

as an aid to explore documents and other data sources for existing causal
knowledge. This can then be used to generate candidate causal graphs, which, while
still requiring human validation or completion, may reduce time-consuming manual
work.

■ The next step in AI requires causal AI. Current deep learning models, in particular
large language models (LLMs) and “reasoning” models for GenAI and AI agents,
have limitations in terms of reliability. A composite AI approach that complements
for example LLMs with causal AI — in particular, causal knowledge graphs — offers
a promising avenue to bring AI to a higher level.

Gartner, Inc. | G00828523 Page 22 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Causality is not trivial. Not every phenomenon is easy to model in terms of its
causes and effects, with many factors potentially being relevant. Causality might be
delayed, circular, unknown or hard to validate, despite the growing use of AI for
causal discovery.

■ The quality of a causal AI model depends on its causal assumptions and on the
data used to build it. This data is susceptible to bias and imbalance and may be
incomplete in terms of representing all causal factors, known or unknown.

■ Causal AI requires technical and domain expertise to properly estimate causal

effects. Building causal AI models is often more difficult than building correlation-
based predictive models, requiring active collaboration between domain experts and
AI experts.

■ AI experts might be unaware of causality methods. If AI experts are overly reliant on

data-driven models like machine learning (ML) or LLMs, organizations could get
pushback when looking to implement causal AI.

■ Limited experience with enterprise-scale applications. This represents a challenge

when organizations run initial causal AI pilots and then want to scale them up to
larger and possibly more complex causal models.

Gartner, Inc. | G00828523 Page 23 of 108

This research note is restricted to the personal use of [email protected].

 User Recommendations

■ Apply causal AI to replace or complement existing AI approaches, including machine

learning, generative AI and agentic AI, to achieve greater reliability and transparency.

■ Use causal AI when more augmentation and automation is required. Examples

include decision intelligence use cases in customer retention programs, marketing
campaign allocation and financial portfolio optimization, as well as in smart
robotics and autonomous systems.

■ Select different causal AI techniques depending on the complexity of the specific use
case. These include ML or LLMs for causal discovery, causal rule inferencing, causal
graphs, Bayesian networks or simulation.

■ Educate your AI teams on causal AI. Explain the difference between causal and
correlation-based AI and cover the range of techniques available to incorporate
causality.

■ Closely involve domain experts in causal AI initiatives to help create, maintain or at

least validate causal models.

Sample Vendors

Actable AI; Bayes Server; causaLens; Causality Link; Geminos; Howso; Parabole.ai;
Scalnyx; Vizuro; Xplain Data

Gartner Recommended Reading

AI Design Patterns for Composite AI

When Not to Use Generative AI

Innovation Insight: AI Agents

Embodied AI
Analysis By: Pieter den Hamer

Benefit Rating: Transformational

Market Penetration: Less than 1% of target audience

Maturity: Embryonic

Gartner, Inc. | G00828523 Page 24 of 108

This research note is restricted to the personal use of [email protected].

 Definition:

Embodied AI is based on the view that intelligence and embodiment in a certain context
are inextricably linked — one shapes the other. It is an approach where a physical or virtual
AI agent’s models are trained and co-engineered with its embodiment: the user interface,
sensors, appearance, actuators or other capabilities required to perceive and interact with
a specific, real or simulated environment. This enables more robust, resilient and adaptive
execution of intelligent tasks.

Why This Is Important

Embodied AI aims to create AI agents that can act autonomously or augment humans in
practical, dynamic contexts — much more so than current AI, including abstract large
language and “reasoning” models with limited reliability and effectiveness in decision
making and action taking. This is achieved through active perception and adaptive
behavior, orchestrated by an AI agent’s intelligence that is in symbiosis with the
capabilities and constraints of the AI agent’s host or body in a certain environment.

Business Impact

Embodied AI paves the way toward more robust, trustworthy, adaptive and actionable AI,
widening its applicability and value creation. This is particularly the case where there is a
need for more practical know-how, physical common sense, social and emotional
intelligence, and a greater resilience to deal with the dynamics and unexpected events in
real-world or virtual environments. Example use cases include autonomous vehicles and
smart robots, but also virtual assistants or gaming characters.

Gartner, Inc. | G00828523 Page 25 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ Recent advances in GenAI and agentic AI are impressive, yet AI still has significant
limitations, particularly with respect to its reliability in dealing with the dynamics and
complexity of reality.

■ Advances in realistic 3D/4D simulations, virtual/augmented/mixed reality and

gaming. Combined with reinforcement learning for adaptive behavior training, this
allows the co-evolving of baseline versions of both embodiment and intelligence of
AI agents, before further deploying and improving them in a real environment, be it
physical or virtual.

■ Emerging approaches include world models, physics-informed or first-principles AI

(representing, among others, the laws of physics or engineering heuristics), adaptive
AI (learning during operations), emotion AI (understanding and expressing feelings in
a social context), composite AI (e.g., using neurosymbolic AI for spatiotemporal
reasoning) and causal AI (representing cause-and-effect relations).

■ Innovation is ongoing in sensor technology, robotics engineering and, for example,

new materials for more natural mechanics and haptic interfaces (relevant for
embodied AI in physical contexts).

■ Scientific insights about intelligence are evolving; intelligence is no longer seen as a

centralized brain-only concept. Cognitive traits like perception, emotion, reasoning
and behavior are often distributed and co-evolved in multiple parts of the body.

■ Investments are being made in research to develop future artificial general

intelligence, for which embodied AI is increasingly seen as a critical step, based on
the view that intelligence is inseparable from its operational entity that interacts with
the environment. This means it is not abstracted from but grounded in reality by
design, holding the promise of providing intrinsic meaning or semantics to its
knowledge representations and “native” common sense.

Gartner, Inc. | G00828523 Page 26 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ The world is a very complex, unpredictable and even chaotic place. That is why the
development of realistic simulations, effective robotics and — for example — truly
autonomous cars has proven to be elusive.

■ Real-world interaction requires real-time, highly responsive AI, even with limited
energy and compute resources (e.g., on mobile or edge devices). However, more
lightweight and energy-efficient AI are not easily achievable.

■ Embodied AI holds the promise of more autonomous AI. Unfortunately, this may not
only facilitate benevolent but also malevolent use. Effective regulation and risk
management for responsible AI are, however, not a given.

■ AI embodiments can be — depending on the use case — unnecessarily humanoid in

their design, bringing in additional complexity and challenges.

■ Embodied AI requires multidisciplinary collaboration between experts in areas as

diverse as machine learning, GUI design and mechanical engineering.

User Recommendations

■ Identify use cases that may benefit from applying embodied AI, both in more virtual
domains, such as online customer interaction or knowledge worker augmentation,
and in more physical domains, such as manufacturing, logistics, healthcare or
facility management.

■ Explore the value that embodied AI can add by reducing the limitations of current AI
in terms of better interpretation of, for example, physical constraints in a warehouse
or cultural norms in client interaction. This may result in increased safety or
decreased bias in the use of AI, respectively.

■ Extend the mindset of how AI agents should be developed or trained. Move from a
modeling-only approach toward one that considers how intelligence can be a
synergy between AI models and the design of the agent’s embodiment. This could,
for example, relate to the facial expression of virtual agents, or the coordination of
movement in physical agents.

Sample Vendors

DEEP Robotics; Figure AI; Guerrilla Games; Intrinsic; NVIDIA; Qualcomm; Sereact; Toshiba;
Unitree; Wayve

Gartner, Inc. | G00828523 Page 27 of 108

This research note is restricted to the personal use of [email protected].

 Gartner Recommended Reading

Innovation Insight: AI Agents

Innovation Insight: AI Simulation

Hype Cycle for Mobile Robots and Drones, 2024

Top Strategic Technology Trends for 2025: Polyfunctional Robots

AI Simulation
Analysis By: Leinar Ramos, Jim Hare, Anthony Mullen

Benefit Rating: High

Market Penetration: 1% to 5% of target audience

Maturity: Adolescent

Definition:

AI simulation is the combined application of AI and simulation technologies to jointly

develop AI agents and the simulated environments in which they can be trained, tested
and sometimes deployed. It includes both the use of AI to make simulations more efficient
and useful, and the use of a wide range of simulation models to develop more versatile
and adaptive AI systems.

Why This Is Important

Increased complexity in decision making is driving demand for both AI and simulation.
However, current AI faces challenges, as it is brittle to change and usually requires a lot of
data. Conversely, realistic simulations can be expensive and difficult to build and run. To
resolve these challenges, a growing approach is to combine AI and simulation: Simulation
is used to make AI more robust and compensate for a lack of training data, and AI is used
to make simulations more efficient and realistic.

Business Impact

AI simulation can bring:

Gartner, Inc. | G00828523 Page 28 of 108

This research note is restricted to the personal use of [email protected].

 ■ Increased value by broadening AI use to cases where data is scarce, private or
confidential, using simulation to generate synthetic data (for example, synthetic data
for autonomous driving training)

■ Greater efficiency by leveraging AI to decrease the time and cost to create and use
complex and realistic simulations

■ Greater robustness by using simulation to generate diverse scenarios, increasing AI

performance in uncertain environments

■ Decreased technical debt by reusing simulation environments to train future AI

models

Gartner, Inc. | G00828523 Page 29 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ Limited availability of AI training data is increasing the need for synthetic data
techniques, such as simulation. Simulation techniques, like physics-based 3D
simulation, are uniquely positioned to generate diverse AI training datasets.
Simulation is able to generate diverse “corner case” scenarios that do not appear
frequently in real-world data, but that are still crucial to train and test AI.

■ Advances in capabilities are making simulation increasingly useful for AI.

Simulation capabilities have been rapidly improving, driven both by increased
computing performance and more efficient techniques.

■ Research in learned simulations (known as “world models”) is driving interest in AI

simulation. Research is increasing on training world models that can learn to predict
how the environment will evolve, based on its current state and agents’ actions.
These learned simulations could make AI simulation more feasible by not having to
directly specify simulation parameters.

■ The emergence of embodied AI is increasing the need for AI simulation. Simulation

environments are often the primary way to train embodied AI (AI adapted to its
physical or virtual context) via reinforcement learning. The increased interest in
embodied AI systems, like robots, is driving AI simulation demand.

■ Increased technical debt in AI drives the need for the reusable environments that
simulation provides. Organizations will increasingly deploy hundreds of AI models,
which requires a shift in focus toward building persistent, reusable environments
where many AI models can be trained, customized and validated. Simulation
environments are ideal since they are reusable, scalable and enable the training of
many AI models at once.

■ The growing sophistication of simulation drives the use of AI, making it more
efficient. Modern simulations are resource-intensive. This is driving the use of AI to
accelerate simulation, typically by employing AI models that can replace parts of the
simulation without running resource-intensive, step-by-step numerical computations.

Gartner, Inc. | G00828523 Page 30 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles
■ Gap between simulation and reality: Simulations can only emulate — not fully
replicate — real-world systems. This gap will reduce as simulation capabilities
improve, but it will remain a key factor. Given this gap, AI models trained in
simulation might not have the same performance once they are deployed;
differences in the simulation training dataset and real-world data can impact
models’ accuracy.

■ Complexity of AI simulation pipelines: The combination of AI and simulation

techniques can result in more complex pipelines that are harder to test, validate,
maintain and troubleshoot.

■ Limited readiness to adopt AI simulation: A lack of awareness among AI

practitioners about leveraging simulation capabilities can prevent organizations
from implementing an AI simulation approach. There will also be skepticism of the
quality and accuracy of simulations, limiting potential adoption.

■ Fragmented vendor market: The AI and simulation markets are fragmented, with
few vendors offering combined AI simulation solutions, potentially slowing down the
deployment of this capability.

User Recommendations

■ Complement AI with simulation to optimize (business) decision making or to

overcome a lack of real-world data by offering a simulated environment for synthetic
data generation or reinforcement learning.

■ Complement simulation with AI by applying deep learning to accelerate simulation,

and generative AI to augment simulation by creating realistic content for
simulations, including images, videos and text. In addition, AI agents can be
embedded in a simulation to make it more realistic.

■ Create synergies between AI and simulation teams, projects and solutions to enable
a new generation of more adaptive solutions for ever-more-complex use cases.
Incrementally build a common foundation of more generalized and complementary
models that are reused across different use cases, business circumstances and
ecosystems.

■ Prepare for the combined use of AI, simulation and other relevant techniques — such
as graphs, natural language processing or geospatial analytics — by prioritizing
vendors that offer platforms that integrate different AI techniques (composite AI) as
well as simulation.

Gartner, Inc. | G00828523 Page 31 of 108

This research note is restricted to the personal use of [email protected].

 Sample Vendors

Altair; Ansys; The AnyLogic Company; Cosmo Tech; Epic Games; MathWorks; Microsoft;
NVIDIA; Rockwell Automation; Unity

Gartner Recommended Reading

Innovation Insight: AI Simulation

World Models
Analysis By: Mike Fang, Nick Ingelbrecht, Sushovan Mukhopadhyay

Benefit Rating: Transformational

Market Penetration: Less than 1% of target audience

Maturity: Emerging

Definition:

World models are learned abstract representations of an environment. They enable AI

systems to make predictions via simulating potential future states and helping to
understand the consequences of the actions taken.

Why This Is Important

AI systems struggle to function effectively in physical environments due to challenges

such as safety concerns, restricted data coverage, limited adaptivity to novel situations,
and the absence of cause-and-effect reasoning capabilities. World models are
fundamental for efficiently forming representations based on the environment,
constructing plans, and simulating events and their outcomes. They offer insights into
potential effects of actions in the environment, which are crucial for AI agents.

Business Impact

■ By capturing the underlying principles and regularities of the environment, world

models can enable the simulation and anticipation of future states and outcomes
based on current conditions and actions. This allows AI systems to acquire
knowledge, refine their models and apply learned insights to new situations for
informed decision making, even in unfamiliar contexts.

■ World models could provide AI applications with a controlled environment for

experimentation, allowing researchers and developers to explore different strategies,
algorithms and policies before deploying them in the real world.

Gartner, Inc. | G00828523 Page 32 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ World models have applicability across various sectors, from film production to
autonomous vehicles and robotics. Their ability to enable simulation and anticipate
complex interactions makes them invaluable tools for AI agents to achieve
innovation and efficiency in diverse fields.

■ World models empower AI to perform more sophisticated prediction and planning

tasks, moving beyond mere pattern recognition in observed data. By simulating and
understanding the dynamics of environments, AI can better handle uncertainty or
missing information and therefore make informed decisions that account for future
possibilities and contingencies.

■ These models can be used to enhance the realism and credibility of generated video
content by incorporating physical laws and constraints. This ensures that the
produced visuals adhere to the principles of physics, resulting in more believable and
immersive experiences.

■ Trained on extensive multimodal datasets derived from robots functioning in real-

world scenarios while combining first-principle AI capabilities, world models can
guide robots in object manipulation and interaction with their environments.

■ World models assist embodied AI in comprehending associations, counterfactuals,

interactions and modeling the dynamics of the world. They go beyond summarizing
observed content by efficiently simulating potential scenarios to predict outcomes,
thereby enabling the selection of optimal actions.

■ Research from control theory and cognitive science, such as Joint-Embedding

Predictive Architecture (JEPA), has highlighted alternative approaches to construct
world models.

Gartner, Inc. | G00828523 Page 33 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Simulating real-world environments and inferring causal relationships is one of the

most challenging domains of AI, and therefore building world models.
Counterfactual reasoning requires simulating hypothetical causes and predicting
outcomes, but current models are limited.

■ Simulating physical laws is challenging for world models, especially in capturing

real-world physical rules. Existing synthetic video generation models like Sora
simulate phenomena like object motion and light reflections, but struggle with
complex physical effects like fluid dynamics and aerodynamics, lacking accuracy
and consistency.

■ Techniques supporting world models mainly interpolate data, not extrapolate. The
real world has many uncertainties, making world models difficult to build.

■ Unlike humans, world models need a very large amount of situational and contextual
combinations, leading to high computational costs. Additionally, acquiring real-world
data faces challenges like public availability, security and privacy issues.

User Recommendations

■ Avoid relying solely on GenAI techniques for world modeling as a solution for every
use case; instead, leverage a broad array of methods from both physical AI and
cognitive science to create a more comprehensive and effective model.

■ Utilize extensive multimodal datasets, including sensory inputs like images and
sounds, to train or customize world models for better contextual understanding and
decision making across diverse scenarios.

■ Implement strategies to mitigate bias and ethical issues in world models, ensuring
fair and unbiased decision making in AI systems.

■ Manage expectations around these techniques, as they are still surrounded by hype.
Begin by piloting them in more focused or non-mission-critical use cases through
mini world models limited to constrained environments (like in game situations).

Sample Vendors

Covariant; Decart; Google; Meta; NVIDIA; OpenAI; VERSES; World Labs

Gartner Recommended Reading

Innovation Insight: Causal AI

Gartner, Inc. | G00828523 Page 34 of 108

This research note is restricted to the personal use of [email protected].

 Innovation Insight: AI Simulation

Emerging Tech: Generative AI’s Path to Revolutionary Computer Vision Products

Decision Intelligence
Analysis By: David Pidsley, Pieter den Hamer, Erick Brethenoux

Benefit Rating: Transformational

Market Penetration: 5% to 20% of target audience

Maturity: Emerging

Definition:

Decision intelligence (DI) is a practical discipline that advances decision making by

explicitly understanding and engineering how decisions are made, and how outcomes are
evaluated, managed and improved via feedback. By digitizing and modeling decisions as
assets, DI bridges the insight-to-action gap to continuously improve decision quality,
actions and outcomes. DI is technology-agnostic and applies decision-centric frameworks
like observe, orient, decide and act (OODA) and Gartner DI (GDI).

Why This Is Important

Agentic AI and GenAI hype, regulatory pressures on decision automation, and recent
global crises have exposed the fragility of business processes and the predigital, implicit
and suboptimal ways of decision making that remain incumbent. DI is positioned beyond
the Innovation Trigger, poised to address these challenges by making decisions more
explicit, optimal, adaptable and auditable.

Gartner, Inc. | G00828523 Page 35 of 108

This research note is restricted to the personal use of [email protected].

 Business Impact

■ Faster, higher quality decisions that are consistent, compliant and cost-effective
while being complex, contextual and continuous, thus driving agility in facing
opportunities and threats in domains like banking, healthcare and supply chain.

■ Enduring, effective, efficient, explainable and ethical enterprisewide DI execution

enhances timely stakeholder outcomes.

■ Risk is mitigated through accurate, trustworthy, fair, privacy-protective and scalable

decision-centric operationalization of AI to augment and automate decisions.

■ Adaptability of decisions-as-assets strengthens decision governance and outcome

predictability.

Gartner, Inc. | G00828523 Page 36 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ Dynamic business complexity: Unpredictable disruptions, chaotic environments and

accelerating pace of digital competition demands near real-time decision models
that can adapt. Decision services can be powered by the composition of multimodal
data analysis, data science, optimization, expert knowledge and other AI techniques.

■ Decision silos: DI curtails fragmented, localized and implicit decisions that

undermine organizational efficacy and efficiency. It also addresses the demand for
cross-functional alignment on decisions-as-assets, the need for harmonization on
which action should be taken following a business decision, and outcome
optimization that balances global efficiency and local adaptations.

■ Deluge of dashboards not driving action: Despite proliferation of “data-driven” tools

and interfaces, most of which fail to connect insights to actions, dashboard
development delays create decision latency, ambiguous outcomes and inability to
perceive a decision’s impact harming organization efficiency.

■ Human-AI delegation and distrust: AI adoption requires transparent, auditable

decision models to address ethical concerns and ensure accountability. Automating
human decisions has promoted disquiet and requires monitoring.

■ Regulatory scrutiny: Data protection, AI and socio-environmental mandates compel

explicit decision documentation for tighter compliance, risk awareness and
mitigation. Explicit decision modeling and decision stewardship drive the analysis,
management and control of the operational processes and observations needed to
enforce decision governance policies and standards applied to decisions-as-assets.

■ Availability and innovation of enabling technologies: Convergence of rule engines,

simulation and optimization in DI platforms practically enables DI prototypes and
pilots to become scalable DI implementations.

■ GenAI acceleration of DI: Enriched context awareness via LLMs is accelerating

composite AI model development for low-code/no-code business decision analysts
and pilots of agentic decision automation.

Gartner, Inc. | G00828523 Page 37 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Business stakeholder apathy, limited urgency and low cultural readiness, ineffective
change management, and lack of DI skills and AI literacy hinder adoption.

■ Bridging the insight-to-action gap to improve outcomes requires a decision-centric

vision beyond the data-driven dogma and the data-to-insight workflow. Technology
centricity overlooks psychological and sociological factors in decision making.

■ Weak collaboration, inadequate operating, delivery and organizational models (i.e., a

DI center of excellence), and disconnected decision-making silos hamper DI
effectiveness. Even advanced cross-silo DI practitioners struggle to impartially
reconsider key decision flows.

■ Unselective or overly enthusiastic adoption of decision automation introduces risks,

including unintended consequences, loss of context and bias amplification. This
undermines trust in DI and limits effective use of DI platforms.

User Recommendations

■ Define and model critical decisions involving resource allocation, uncertainty or

competing alternatives. Use these as pilots to build DI momentum and demonstrate
value for enterprisewide adoption.

■ Inventory repetitive, high-impact decisions and their key inputs. Adopt decision-
centric modeling by articulating outcomes, decision logic, alternative courses of
actions and required observations to drive continuous learning, improvement and
transparency.

■ Maximize decision quality, resilience and traceability through cross-functional DI

fusion teams, fostering collaboration and alignment across departments. Delegate
decision-making authority to those with the most relevant expertise and context.

■ Upskill staff in decision modeling, prescriptive analytics and optimization.

Investigate the roles of decision engineers, decision scientists and decision
stewards. Experiment with agentic, GenAI and other composite AI, and DecisionOps
to support organizationwide decision centricity and excellence.

Gartner Recommended Reading

Market Guide for Decision Intelligence Platforms

Innovation Insight: Decision Intelligence for Chinese Organizations

Gartner, Inc. | G00828523 Page 38 of 108

This research note is restricted to the personal use of [email protected].

 Quick Answer: Where to Reengineer Your Business Decisions

How to Manage the Risks of Decision Automation

When to Automate or Augment Decision Making

FinOps for AI
Analysis By: Jim Hare, Adam Ronthal, Andrei Razvan Sachelarescu

Benefit Rating: High

Market Penetration: 1% to 5% of target audience

Maturity: Emerging

Definition:

FinOps for AI is the application of financial operations best practices to help

organizations increase visibility and manage the costs of AI services to ensure efficient
usage and deliver maximum business impact. Using FinOps to track and measure AI
spend and usage is crucial for optimizing costs, ensuring financial accountability and
maximizing ROI.

Why This Is Important

Cost poses one of the greatest near-term threats to AI and GenAI success. AI workloads,
especially in cloud environments, often use expensive GPU-based compute infrastructure
and consume tokens in unforeseen ways, leading to unpredictable expenses if not
monitored properly. Deploying and managing AI solutions generates other costs, including
development, governance and change management. Using FinOps to track and measure
AI spend and usage is crucial for optimizing costs, ensuring financial accountability and
maximizing ROI.

Gartner, Inc. | G00828523 Page 39 of 108

This research note is restricted to the personal use of [email protected].

 Business Impact

FinOps helps businesses optimize AI spend by providing real-time cost visibility and
control, enabling teams to allocate resources efficiently and prevent budget overruns while
also preventing underprovisioning that can cause downtime or slowdowns. FinOps for AI
also enhances collaboration between finance, engineering and operations teams, ensuring
that AI investments align with business objectives while ensuring cost-efficiency. Using
FinOps practices, organizations can maximize the ROI of AI initiatives and leverage cost-
saving opportunities such as reserved instances, workload automation/optimization and
usage-based pricing models.

Drivers
■ AI adoption, especially AI applications and GenAI, is contributing to a spike in cloud
costs for most enterprises. Hidden costs and unpredictable invoices make it difficult
for organizations to deploy AI more broadly.

■ Tracking AI costs and usage scaling is complex due to fluctuating computational

demands, variable AI service pricing, hidden infrastructure costs and exponential
scaling of model training and inference across users and applications.

■ Organizations new to AI and/or the cloud are unlikely to be prepared for AI cost
volatility and will need to adjust their legacy operating models and budget practices
by adopting FinOps for AI. Many organizations face challenges in tracking and
measuring AI costs against concrete business benefits.

■ Engineering teams are often immature in their use of AI services and the many
dynamic layers needed to achieve ongoing cost-effectiveness.

■ The total cost of ownership (TCO) of AI use cases can differ from the cost of
traditional software applications with fixed costs and purpose. Continuous training,
switching to newer models, specialized infrastructure like GPUs and differences in
processing costs for specific data types (text, image, video, audio) are part of
ongoing AI costs.

■ Many AI models and services are based on consumption pricing models and may be
purchased in many versions or variants.

■ Pricing may also fluctuate based on a variety of factors such as usage, model
choice, accuracy and performance guarantees. The velocity of pricing volatility
requires continuous and active assessments of price/performance and accuracy.

Gartner, Inc. | G00828523 Page 40 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Implementing FinOps for AI is challenging because of AI workloads’ unpredictable

and dynamic nature and the complexity and variety of cost factors that make cost
estimation, budgeting and optimization more challenging compared with traditional
cloud operations.

■ Balancing performance and cost-efficiency is difficult because AI models often

require specialized compute infrastructure resources, GPUs and large datasets that
can lead to excessive cloud spending if they are not monitored and optimized
effectively.

■ Many organizations struggle with cross-functional collaboration among finance,

operations and engineering teams, as aligning AI-specific cost insights with business
objectives requires a cultural shift and enhanced visibility into AI-driven
expenditures.

Gartner, Inc. | G00828523 Page 41 of 108

This research note is restricted to the personal use of [email protected].

 User Recommendations

■ Track the TCO of using AI: Implement real-time tracking of total running costs,
including cloud, infrastructure and labor costs. Use tagging and cost allocation
strategies to assign expenses to specific AI projects or departments. Assign budgets
to AI-related resource and service groups, and trigger cost alerts when consumption
exceeds budget goals.

■ Optimize AI spend and workloads: Track AI spend across packaged and custom
SaaS, AI-leveraging commercial models (tokens via API calls), and compute from
hosted models.

■ Understand the pros and cons of buying versus building models: Closed models
built by model providers may be considered more expensive at first glance, but they
reduce delivery time, upfront development costs and the need for more expensive
skills. Build models for truly strategic types of use cases.

■ Embrace an agile approach to model switching: Regularly compare models in use

with alternative options to see whether the same or better accuracy can be achieved
at lower cost.

■ Invest in making data AI-ready: Control data preparation and processing costs by
investing in data cleansing and curation to produce smaller training and retrieval-
augmented generation datasets of higher quality.

■ Implement proactive cost management controls and guardrails: Integrate real-time

anomaly detection and alerts with demand-throttling options to guard against
unexpected cost spikes.

Sample Vendors

Airia; Exostellar; FinOps Foundation; Finout; Flexera; IBM

Gartner Recommended Reading

Beyond FinOps: Optimizing Your Public Cloud Costs

The State of FinOps for Data and Analytics, 2024

Emerging Tech: Data Management Solutions Need Augmented FinOps

Key Data and Analytics Platform Insights for FinOps Providers

Gartner, Inc. | G00828523 Page 42 of 108

This research note is restricted to the personal use of [email protected].

 Data and Analytics Essentials: FinOps and Cloud Operating Models

Neurosymbolic AI
Analysis By: Erick Brethenoux, Afraz Jaffri

Benefit Rating: High

Market Penetration: 5% to 20% of target audience

Maturity: Emerging

Definition:

Neurosymbolic AI is a form of composite AI that combines probabilistic reasoning

methods and symbolic systems to create more robust and trustworthy AI models. This
fusion enables the combination of probabilistic models with logic-based techniques (such
as rules and knowledge graphs) to enable AI systems to better represent, reason and
generalize concepts. This approach provides a reasoning infrastructure for solving a wider
range of business problems more effectively.

Why This Is Important

Neurosymbolic AI addresses limitations in current AI systems, such as incorrect outputs,

lack of generalization to a variety of tasks and an inability to explain the steps that led to
an output. The neurosymbolic approach leads to more powerful, versatile and
interpretable AI solutions and allows AI systems to reason through more complex tasks.
Generative AI systems are starting to leverage neurosymbolic methods to overcome their
reasoning shortcomings.

Business Impact

Neurosymbolic AI will have an impact on the efficiency, adaptability and reliability of AI

systems used across business processes. The integration of logic and multiple reasoning
mechanisms brings down the need for ever larger AI models and their supporting
infrastructure. These systems will rely less on the processing of huge amounts of data,
making AI agile and resilient. Neurosymbolic approaches can augment and automate
decision making with less risk of unintended consequences.

Gartner, Inc. | G00828523 Page 43 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ Neurosymbolic AI addresses the limitations of large reasoning models (LRMs),

which are still plagued with a lack of symbolic abstraction when exclusively based
on deep learning techniques.

■ The need for explanation and interpretability of AI outputs is especially important in

regulated industry use cases and in systems that use private data.

■ Understanding the meanings behind words, not just their arrangement (semantics
over syntax), is an increasing priority in systems that deal with real-world entities to
ground meaning to words and terms in specific domains.

■ The set of tools available to combine different types of AI models is increasing and
becoming easier to use for developers and end users. The dominant approach is to
chain together results from different models (composite AI) rather than using single
models.

■ The integration of multiple reasoning mechanisms necessary to provide agile AI

systems eventually leads to adaptive AI systems, notably through blackboardlike
mechanisms.

■ Agentic AI advances also participate in advancing neurosymbolic methods, while

agents using various composite AI techniques collaborate to solve problems.

Obstacles

■ Most fundamental neurosymbolic AI methods and techniques are being developed

in academia or industry research labs. Despite the increased availability of tools,
implementations in business or enterprise settings are still limited.

■ No agreed-upon techniques exist for implementing neurosymbolic AI, and

disagreements continue between researchers and practitioners on the effectiveness
of combining approaches, despite the emergence of real-world use cases.

■ The commercial and investment trajectories for AI startups allocate almost all
capital to deep-learning approaches, leaving only those willing to bet on the future to
invest in neurosymbolic AI development.

■ Currently, despite increasing exposure, popular media and academic conferences do

not give as much exposure to the neurosymbolic AI movement as compared to other
approaches (such as generative AI).

Gartner, Inc. | G00828523 Page 44 of 108

This research note is restricted to the personal use of [email protected].

 User Recommendations

■ Adopt composite AI approaches when building AI systems by using a range of

techniques that increase the robustness and reliability of AI models. Neurosymbolic
AI approaches will fit into a composite AI architecture.

■ Dedicate time to learning about neurosymbolic AI approaches, and to identifying use

cases that can benefit from applying these approaches.

■ Invest in data architecture that can leverage the building blocks for neurosymbolic AI
techniques, such as knowledge graphs and agent-based techniques.

■ Consider neurosymbolic AI architectures when the limitations of generative AI

models prevent their implementation in the organization.

■ Educate developers on the potential of neurosymbolic models by exploring the

capabilities of neurosymbolic approaches while building learning AI agents.

Sample Vendors

Franz; Google DeepMind; IBM; Microsoft; RelationalAI; Wolfram|Alpha

Gartner Recommended Reading

Innovation Insight: AI Simulation

Go Beyond Machine Learning and Leverage Other AI Approaches

When Not to Use Generative AI

AI Zodiac: Mapping AI Use Cases to Techniques

Artificial General Intelligence

Analysis By: Pieter den Hamer, Philip Walsh

Benefit Rating: Transformational

Market Penetration: Less than 1% of target audience

Maturity: Embryonic

Gartner, Inc. | G00828523 Page 45 of 108

This research note is restricted to the personal use of [email protected].

 Definition:

Artificial general intelligence (AGI) is the (currently hypothetical) capability of a machine

that can match or surpass the capabilities of humans across all cognitive tasks. In
addition, AGI will be able to autonomously learn and adapt in pursuit of predetermined or
novel goals in a wide range of both physical and virtual environments.

Why This Is Important

With AI’s growing sophistication — including the recent advances in generative AI (GenAI)
and agentic AI — a growing number of AI experts have shortened their predicted timelines
for achieving AGI in the future or view AGI as no longer purely hypothetical. A clear, shared
definition of AGI is necessary for evidence‑based governance and realistic expectations.
Achieving AGI would be a transformative tipping‑point with profound consequences for
productivity, employment, geopolitical power, legal, ethical and cultural norms — and
society at large.

Business Impact

In the near term, anticipation of AGI fuels both overly optimistic expectations and
existential fears, skewing investment, distorting trust and accelerating the emergence of
new AI regulations. Over the longer horizon, the question of who builds and controls AGI —
or other forms of increasingly powerful AI — looms large. Many experts see public
stewardship as essential, a prospect that could upend private advantage and redraw
entire markets.

Gartner, Inc. | G00828523 Page 46 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ Recent advances and growing interest in multimodal large language models (LLMs),
so-called reasoning models and AI agents drive considerable hype about AGI. The
massive scaling of deep learning and the availability of huge amounts of data and
compute power largely have enabled these advances.

■ AI’s further evolution toward AGI, as defined here, is increasingly complemented by

other partially new approaches, such as knowledge or causal graphs, world models,
adaptive AI, embodied AI, composite and neurosymbolic AI, and likely other
innovations yet unknown.

■ A number of AI vendors are openly discussing and actively researching the field of
AGI, creating the impression that AGI lies within reach. However, their definitions of
AGI vary greatly and are often open to multiple interpretations. Moreover, other
leading AI vendors and experts have dismissed AGI as hype, urging focus on the real
impact of AI’s growing capabilities.

■ Humans’ innate desire to set lofty goals is also a major driver for AGI. At one point in
history, humans wanted to fly by mimicking bird flight. Today, airplane travel is a
reality. The inquisitiveness of the human mind, taking inspiration from nature and
from itself, is not going to fizzle out.

■ People’s tendency to anthropomorphize nonhuman entities also applies to AI-

powered machines. The humanlike responses of LLMs and the reasoning-like
capabilities of recent AI models have fueled this tendency. Although many
philosophers, neuropsychologists and other scientists consider this attribution highly
uncertain or going too far, it has created a sense that AGI is within reach or at least is
getting closer. In turn, this has triggered massive media attention, several calls for
regulation to manage the risks of AGI and a great appetite to invest in AI for
economic, societal and geopolitical reasons.

Gartner, Inc. | G00828523 Page 47 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Little scientific consensus exists on the meaning of “human intelligence.” Any claims
about AGI are hard to validate in the face of the enormous complexity of the human
brain and mind and such a limited understanding of them.

■ Unreliability, lack of transparency and limited abstraction and reasoning of pattern-

based capabilities in current AI are not easy to overcome with deep learning’s
intrinsically probabilistic approach. More data or more compute power for ever-
bigger models is unlikely to resolve these issues, let alone to achieve AGI. To realize
(and control) AGI will require further technological innovations. Therefore, AGI as
defined here is unlikely to emerge in the near future.

■ If AGI materializes, autonomous actors likely will emerge that, in time, will be
attributed with full self-learning, agency, identity and perhaps even morality. This will
open a bevy of considerations about AI’s legal rights and trigger profound ethical
and even religious discussions. AGI also brings the risk of negative impacts on
humans, from job losses to a new, AI-triggered arms race and more. A serious
backlash may result, and regulations to ban or control AGI are likely to emerge in the
near future.

User Recommendations
■ Engage with stakeholders to address excessive optimism or unwarranted
pessimism, and create or maintain realistic expectations around AGI. Ground AI
strategy in concrete business problems rather than speculative AGI forecasts.
Recalibrate the AI portfolio periodically as AI capabilities evolve, while leveraging the
complementary strengths of human and artificial intelligence.

■ Stay apprised of scientific and innovative breakthroughs that may indicate AGI’s
possible emergence; however, be aware of the broad range of definitions and views
regarding AGI, some strict and some less strict. Meanwhile, keep applying current AI
to learn, reap its benefits and develop practices for its responsible use.

■ Assess whether AI systems truly meet their specific use-case needs, rather than
relying on generic measures of intelligence.

■ Prepare for emerging AI regulations and promote internal AI governance to manage

current and emerging AI risks. Because although AGI as defined here is not a reality
now, current AI already poses significant risks regarding ethics, reliability and other
areas.

Gartner, Inc. | G00828523 Page 48 of 108

This research note is restricted to the personal use of [email protected].

 Sample Vendors

Aigo; Amazon; Anthropic; Butterfly Effect; DeepSeek; Google; Microsoft; OpenAI

Gartner Recommended Reading

Artificial General Intelligence: 5 Future Perspectives and Present Imperatives

Video: Why Artificial General Intelligence (AGI) Matters — and Why It Doesn’t

How to Think About Artificial General Intelligence

Composite AI
Analysis By: Erick Brethenoux, Pieter den Hamer

Benefit Rating: Transformational

Market Penetration: 20% to 50% of target audience

Maturity: Early mainstream

Definition:

Composite AI, also known as hybrid AI, refers to the combined application (or fusion) of
different AI techniques to improve the efficiency of learning and broaden the level of
knowledge representations. It broadens AI abstraction mechanisms and, ultimately,
provides a platform to solve a wider range of business problems effectively.

Why This Is Important

Composite AI recognizes that no single AI technique is a panacea. It combines

“connectionist” AI approaches, like machine learning (ML) and deep learning, with
“symbolic” and other AI approaches, like rule- and logic-based reasoning, graph or
optimization techniques. The goal is to enable AI solutions to generalize and learn,
embodying more abstraction mechanisms. Composite AI is at the center of the generative
AI (GenAI), decision intelligence (DI) platform and agentic AI markets.

Gartner, Inc. | G00828523 Page 49 of 108

This research note is restricted to the personal use of [email protected].

 Business Impact

Composite AI brings the power of AI to a broader group of organizations that do not have
access to large amounts of historical or labeled data but possess significant human
expertise. It helps to expand the scope and quality of AI applications addressing more
types of reasoning challenges. Other benefits include better interpretability, embedded
resilience and the support of augmented intelligence. The new wave of GenAI
implementations heavily relies on composite AI.

Gartner, Inc. | G00828523 Page 50 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ The growing reliance on AI for decision making is driving organizations toward

composite AI. The most appropriate actions can be determined by combining rule-
based and optimization models — a combination often referred to as prescriptive
analytics.

■ Small datasets, or the limited availability of data, have pushed organizations to

combine multiple AI techniques. Enterprises have started to complement scarce raw
historical data with additional AI techniques, such as knowledge graphs and
generative adversarial networks (GANs), to generate synthetic data.

■ Combining AI techniques is much more effective than relying only on heuristics or a

crudely “data-driven” approach. A rule- and logic-based technique can be combined
with a deep learning model. Rules coming from human experts, or the application of
physical/engineering model analysis, may specify that certain sensor readings
indicate inefficient asset operations.

■ Computer vision and natural language processing (NLP) solutions are used to
identify and categorize people or objects in an image. This output can be used to
enrich or generate a graph, representing the image entities and their relationships.

■ The advent of DI platforms and AI agent-based systems will promote the efficient
combination of multiple AI techniques, making composite AI (and in some instances
neurosymbolic techniques) a de facto AI development approach.

■ The development of multiagent systems is further empowering composite AI. A

composite AI solution can be composed of multiple agents, each representing an
actor in the ecosystem. Combining these agents into a “swarm” enables the creation
of common situation awareness, more global planning optimization, responsive
scheduling and process resilience.

■ GenAI is accelerating the research and adoption of composite AI models through

artifacts, process and collaboration generation, which are the foundation of DI
platforms.

Gartner, Inc. | G00828523 Page 51 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles
■ Lack of awareness and skills in leveraging multiple AI methods: This could prevent
organizations from considering the techniques particularly suited to solving specific
problem types.

■ ModelOps deployment: The ModelOps domain (that is, the operationalization of

multiple AI models, such as optimization models, ML models, rule models and graph
models) remains an art much more than a science. A robust ModelOps approach is
required to efficiently govern composite AI environments and harmonize it with other
disciplines, such as DevOps and DataOps.

■ Trust and risk barriers: The AI engineering discipline is starting to take shape, but
only mature organizations apply its benefits in operationalizing AI techniques.
Security, ethical model behaviors, observability, model autonomy and change
management practices must be addressed across the combined AI techniques.

User Recommendations

■ Identify projects in which a crude “data-driven,” ML-only approach is inefficient or ill-

fitting. For example, in cases when enough of the right data is not available or when
the pattern cannot be represented through current ML models.

■ Capture domain knowledge and human expertise to provide context for data-driven
insights by applying decision intelligence with business rules and knowledge graphs,
in conjunction with ML and/or causal models.

■ Combine the power of ML, image recognition or NLP with graph analytics to add
higher-level, symbolic and relational intelligence.

■ Extend the skills of ML experts, or recruit or upskill additional AI experts to cover

graph analytics, optimization or other DI-relevant techniques for composite AI. For
rules and heuristics, consider knowledge engineering skills, as well as emerging
skills such as prompt engineering.

■ Accelerate the development of DI solutions by encouraging experimentation with

agentic and GenAI, which will in turn accelerate the necessity for deployment of
composite AI solutions.

Sample Vendors

ACTICO; Aera Technology; Almawave; FICO; Filuta; Fujitsu; IBM; Indico Data; SAS

Gartner, Inc. | G00828523 Page 52 of 108

This research note is restricted to the personal use of [email protected].

 Gartner Recommended Reading

Go Beyond Machine Learning and Leverage Other AI Approaches

When Not to Use Generative AI

Top Strategic Technology Trends for 2022: AI Engineering

How to Choose Your Best-Fit Decision Management Suite Vendor

Gartner, Inc. | G00828523 Page 53 of 108

This research note is restricted to the personal use of [email protected].

 At the Peak
AI TRiSM
Analysis By: Avivah Litan, Jeremy D'Hoinne, Bart Willemsen, Lauren Kornutick, Max Goss,
Andrew Bales, Sumit Agarwal

Benefit Rating: High

Market Penetration: 5% to 20% of target audience

Maturity: Early mainstream

Definition:

AI trust, risk and security management (TRiSM) comprises four layers of technical
capabilities that support enterprise policies for all AI use cases and help assure AI
governance, trustworthiness, fairness, safety, reliability, security, privacy and data
protection. The top two layers — AI governance, and AI runtime inspection and
enforcement — are new to AI and are, in part, consolidating into a distinct market
segment. The bottom two layers represent traditional technology focused on AI.

Why This Is Important

AI brings new trust, risk and security management challenges that conventional controls
do not address. Enterprises are most concerned with data compromise, third-party risks,
and inaccurate or unwanted outputs, and need to ensure enterprise AI behavior and
actions align with enterprise intent. Enterprises must further retain independence from any
single AI model or hosting provider to ensure scalability, flexibility, cost control and trust,
as AI markets rapidly mature and change.

Business Impact

Organizations that do not consistently manage AI risks are exponentially inclined to

experience adverse outcomes such as project failures, AI misperformance and
compromised data confidentiality. Inaccurate, unethical or unintended AI outcomes,
process errors, uncontrolled biases, and interference from benign or malicious actors can
result in security failures, financial and reputational loss, or liability and social harm. AI
misperformance can also lead organizations to make suboptimal or incorrect business
decisions.

Drivers

Gartner, Inc. | G00828523 Page 54 of 108

This research note is restricted to the personal use of [email protected].

 ■ The increasing use of AI and GenAI is limited by a lack of trust in AI as a safe and
ethical option for supporting critical business processes.

■ Enterprises face multiple AI risks and are most concerned with data compromise,
third-party risks, and inaccurate or unwanted outputs.

■ Malicious hacks against enterprise AI are still uncommon, while incidents of

unconstrained harmful chatbots are well-documented, and internal oversharing data
compromises are prevalent.

■ User demand for GenAI TRiSM solutions is steadily increasing, and providers of all
sizes are competing for this new enterprise business.

■ Some organizations are mostly concerned with security and risk mitigation, while
others also focus on supporting ethical or safe practices and regulatory compliance.

■ AI trust, risk and security issues surface organizational silo issues, pushing teams to
realign to solve problems that cross departmental boundaries and to implement
technical measures that address them.

■ The introduction of LLM-based AI agents with differing degrees of agency allows AI

to take actions with or without human intervention.

■ The rapid proliferation of AI agents will create more need for governance than
human-in-the-loop oversight can fulfill alone.

■ Regulations for AI risk management are driving businesses to institute measures for
managing AI risk. Such regulations define new compliance requirements that
organizations will have to meet on top of existing ones, like those pertaining to
privacy protection.

Obstacles
■ Adopting AI TRiSM technology is often an afterthought. Organizations generally
don’t consider it until AI applications are in production, when it becomes challenging
to retrofit.

■ Many enterprises are resource-constrained and don’t have the skills or capacity to
implement TRiSM technology.

Gartner, Inc. | G00828523 Page 55 of 108

This research note is restricted to the personal use of [email protected].

 ■ Enterprises often rely on their incumbent vendors to provide AI functionality for
TRiSM capabilities though they often lack it, and must rely on vendor licensing
agreements to ensure their confidential data remains private in the host
environment.

■ Off-the-shelf software that embeds AI is often closed and does not support APIs to
third-party AI TRiSM products that can enforce enterprise policies.

■ Most AI threats and risks are not fully understood and not effectively addressed.

■ AI TRiSM requires a cross-functional team, including legal, compliance, security, IT

and data analytics staff, to establish common goals and use common frameworks.

■ While AI TRiSM can integrate life cycle controls, this poses significant
implementation challenges especially with embedded AI that often lacks API
support.

User Recommendations
■ Set up an organizational unit to manage AI TRiSM that is integrated with the
organization’s overarching governance initiative. Include members with a vested
interest in AI projects.

■ Define acceptable use policies at a level granular enough to enforce.

■ Discover and inventory all AI used in the organization, leveraging the capabilities of
AI TRiSM vendors who support this.

■ Revisit and implement data classification, protection and access management

across all enterprise information — both structured and unstructured — that can
potentially be utilized by AI. Collaborate across different teams involved in
information governance.

■ Work with legal and compliance to contractually define accountability and

responsibility for unacceptable AI use or behavior in third-party-embedded AI
applications.

■ Obtain vendor attestation to meet legal requirements.

■ Evaluate and implement layered AI TRiSM technology to continuously support and

enforce policies across all AI entities in use. This includes enterprise-owned AI
TRiSM services, along with controls offered by frontier model providers, but do not
solely rely on the latter.

Gartner, Inc. | G00828523 Page 56 of 108

This research note is restricted to the personal use of [email protected].

 Gartner Recommended Reading

Market Guide for AI Trust, Risk and Security Management

Top Strategic Technology Trends for 2024: AI Trust, Risk and Security Management

The Cybersecurity Leader’s Playbook: Navigating the EU AI Act

Multimodal AI
Analysis By: Nick Ingelbrecht, Sushovan Mukhopadhyay, Yogesh Bhatt

Benefit Rating: Transformational

Market Penetration: 1% to 5% of target audience

Maturity: Adolescent

Definition:

Multimodal AI models are trained with multiple types of data (also known as modalities)
simultaneously, such as images, video, audio and text. This enables them to create a
shared data representation to improve performance in different tasks. At runtime, they can
handle more than one modality, either in their inputs, their outputs or both.

Why This Is Important

Multimodal AI adds significant new technology capabilities such as greater accuracy to

existing software. It spurs new specialized applications, enables new use cases such as
visual question answering of image frames, manufacturing optimization, and fraud
detection in banking and finance, and creates new value outcomes. The physical world
and the data it generates are inherently multimodal. By integrating and analyzing diverse
data sources, a more comprehensive evaluation of complex environments and tasks can
be achieved compared with unimodal models, helping users make sense of the world and
opening up new avenues for AI applications.

Business Impact

Gartner forecasts that:

■ Over the next five years, multimodal AI will become increasingly integral to capability
advancement in every application and software product across all industries.

Gartner, Inc. | G00828523 Page 57 of 108

This research note is restricted to the personal use of [email protected].

 ■ By 2027, 40% of generative AI (GenAI) solutions will be multimodal (text, images,
audio and video), up from 1% in 2023.

Drivers

Multimodal AI adoption will generate cross sector transformational opportunities. Key

drivers include:

■ A paradigm shift from traditional, linear processes to dynamic, AI-driven systems

where humans and machines collaborate seamlessly. And further evolution of
agentic AI will involve increasing integration with multimodal AI techniques to
handle the complexity and richness of real-world data and tasks.

■ Recent AI breakthroughs, particularly in the realm of large language models (LLMs)

and vision language models (VLMs), are highly relevant to multimodal AI. These
advancements have catalyzed a renaissance in natural language processing and
computer vision.

■ Intelligent applications, by their nature, are context-rich and designed to adapt to

constantly changing scenarios. This makes multimodal AI a crucial component for
their development and evolution.

■ World models are a significant driver for multimodal AI because they inherently
require the ability to process and understand information from various modalities to
accurately represent and simulate the complexity of the real world.

■ Broader availability of AI/GenAI multimodal models, both proprietary and open-

source, lowers the barriers to entry and adoption via AI marketplaces.

■ There is a demand for multimodal domain-specialized models in areas such as

healthcare, where multimodality extends or enriches use cases.

Obstacles

Multimodal AI is powerful in understanding and processing from various modalities, but

faces several primary obstacles to adoption:

■ Integrating diverse data types — such as text, images, audio and video — is
challenging due to differences in format and time stamps, risking inaccurate
interpretations. Multimodal AI models are complex, combining various modality-
specific subnetworks, which can obscure transparency and explainability.

Gartner, Inc. | G00828523 Page 58 of 108

This research note is restricted to the personal use of [email protected].

 ■ Architectural complexity, increased data volume and the need for data fusion lead to
inference latency, hindering reliable operation where immediate decision making is
crucial.

■ Dataset bias originates from leveraging training datasets like text, images, videos
and speech, which may inadvertently reflect societal or cultural biases. This can
result in making unfair or inaccurate predictions/decisions.

■ Handling sensitive data across modalities increases breach risks and privacy
violations. This complicates compliance with regulations like General Data
Protection Regulation (GDPR) or Health Insurance Portability and Accountability Act
(HIPAA), as multimodal AI exposes new attack surfaces and heightens privacy risks
with diverse data types.

User Recommendations

Organizations looking to implement multimodal AI should:

■ Identify AI use cases where multimodal AI can enhance business value beyond
unimodal AI foundation models.

■ Run pilots with off-the-shelf multimodal models to demonstrate not only technical
feasibility but the business value.

■ Build a strong model evaluation by assessing the quality of relationships between

modalities such as comparing generated captions from images to ground-truth
labels/descriptions.

■ Prioritize building or accessing robust data infrastructure supporting the collection,

storage and processing of diverse data types (text, images, audio and video).

■ Build or acquire expertise to handle the technical complexities of processing and

integrating multimodal data with legacy and existing workflows.

■ Create or extend AI governance strategies and policies to address challenges with

multimodal datasets and ensure compliance.

■ Incorporate multimodality into technology roadmaps and create migration paths for
multimodal AI in systems procurement or product development plans.

Gartner, Inc. | G00828523 Page 59 of 108

This research note is restricted to the personal use of [email protected].

 Sample Vendors

Aimesoft; Google; Hugging Face; Jina AI; Meta; Midjourney; NVIDIA; OpenAI; Stability AI;
TwelveLabs

Gartner Recommended Reading

Emerging Tech: Generative AI’s Path to Revolutionary Computer Vision Products

Emerging Tech: Multimodal Generative AI Interfaces Transform User Experiences

Innovation Insight: Multimodal AI Explained

Sovereign AI
Analysis By: Lydia Clougherty Jones, Clementine Valayer

Benefit Rating: High

Market Penetration: 20% to 50% of target audience

Maturity: Early mainstream

Definition:

Sovereign AI is the effort by nation-states to invest in and progress their own development
and use of AI to advance their unique sovereign objectives. Given the wide variances
across sovereign AI innovation to risk ratios, sovereign AI impacts international
relationships, global trade, local public/private partnerships and economic markets in
unexpected ways.

Why This Is Important

Sovereign AI reflects the rapid acceleration of nation-states’ investment to advance

adoption of AI techniques for their own use, including to improve alignment of their
internal government functions with operational goals. While sovereign AI could enhance
an individual state’s military defense, AI use by other nations could undermine those
national security efforts. Sovereign AI aims to maximize AI value while decreasing AI risk,
including for those sovereign states that collaborate to achieve common goals such as
decreasing the impact of AI-generated deepfakes in political environments.

Gartner, Inc. | G00828523 Page 60 of 108

This research note is restricted to the personal use of [email protected].

 Business Impact

Sovereign AI impacts nearly all aspects of government and the enterprises with which it
interacts. It improves the effectiveness of operations by automating tasks like those
within government contact centers. Because sovereign AI modernizes the business of
government, it can improve employee experience and accelerate citizen engagement.
When sovereign states control their own AI systems, they reduce their dependence on
other sovereign states and on the private tech market.

Gartner, Inc. | G00828523 Page 61 of 108

This research note is restricted to the personal use of [email protected].

 Drivers
■ An increasing number of countries are actively planning and building their own AI
infrastructure and capabilities to increase their competitiveness and safeguard their
future, including by developing sustainable sovereign AI.

■ Known and unknown risks of harms to sovereign objectives and citizen welfare from
irresponsible uses of AI drive sovereign states to desire greater control over the
development of AI systems, and more so over generative AI (GenAI) use cases.

■ Need for sovereign entities to self-regulate, including how their data is used to train
large language models (LLMs), is growing. For example, nation-states are
increasingly using AI tools to make important government decisions, but clients
report that these decisions are often outsourced to private companies without public
input or oversight. This lack of transparency and accountability drives sovereign
states to develop the AI tools themselves to address concerns about unwanted
biases and conflicts of interest in these critical decision-making processes.

■ Nation-states desire to increase their independence from other nations and the tech
market, especially to remedy underrepresentation of cultural and linguistic inputs.

■ Sovereign AI plays an important role in digital sovereignty as it focuses on the

sovereign control of AI data and systems, including control over computing capacity,
data storage, access to human resources and proprietary knowledge for AI
application development. Digital sovereignty also can significantly impact sovereign
AI development, with the availability of locally stored data, for example, to train AI
models.

■ Sovereign AI differs from sovereign data strategies. Its core focus is being the
developer and user of AI technologies, not the regulator of them. Sovereign data
strategies reflect state efforts to regulate data about and AI use by its citizens,
private industry and its economy.

■ Combating threats to political stability from the proliferation of deepfakes can drive
nation-states to adopt sovereign AI.

■ Upskilling government workers today for a more AI-ready government workforce

tomorrow will assist sovereign AI advancement.

■ Demand to advance local and national defense strategies is increasing.

■ Progressing and maintaining leadership in emerging technologies space will

accelerate scaling AI.

Gartner, Inc. | G00828523 Page 62 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles
■ Preparation of an AI-ready IT infrastructure.

■ Development of an AI-skilled government workforce.

■ Modernization of government culture to embrace advanced analytics and

automation.

■ Limited capability of already-taxed IT infrastructure and fragmented business

networks that would need to serve a sovereign AI system.

■ Lack of the right AI-ready data for training LLMs, resulting in AI output with varying
levels of utility.

■ Lack of technically skilled humans to loop into the AI development and use life cycle,
resulting in an increase of unintended negative outcomes from AI use and GenAI
artifacts.

■ Lack of AI-readiness across people, talent, data management techniques, and

enabling tools and technologies, impeding nation-states from accelerating AI
adoption across government functions.

■ Differences in political and cultural values that will create inconsistent AI-value
versus AI-harms analysis, leading to unpredictable impacts on international trade
and global markets.

■ Legal, regulatory or fiscal policies that can impede sovereign AI initiatives.

■ Development of AI by nation-states across the world that leads to a fragmentation

and possible contradiction in the requirements for AI solutions, many of which
cannot be met by either the public or private sector.

User Recommendations

Sovereign states seeking a self-governed and controlled approach to the development of

AI systems aligned with their strategic objectives should:

■ Start small and prioritize AI uses aligned with maximum advancement of their
stakeholder and government business goals.

■ Build an AI strategic roadmap that progresses from internal use cases to citizen-
facing ones.

Gartner, Inc. | G00828523 Page 63 of 108

This research note is restricted to the personal use of [email protected].

 ■ Ensure the AI strategy identifies key value opportunities and risks.

■ Develop early in the planning process a metric of success and include various
metrics within the strategy to track successes and course-correct unwanted impacts.

■ Cultivate public-private collaborations that foster and accelerate data and analytics,
AI upskilling, technology innovation and adoption.

■ Monitor and learn from sovereign AI already underway, including from New Zealand,
the European Commission, India, the United States, Canada and the United
Kingdom.

■ Collaborate with friendly sovereign states to accelerate the learning curve, sharing
failure analysis and positive narratives of unexpected success.

■ Use Gartner’s The Pillars of a Successful Artificial Intelligence Strategy to guide the
nation toward self-governance of AI development, creating tangible value and
achieving competitive national leader status.

Gartner Recommended Reading

Top Trends in AI Public Policy and Regulations for 2024

The Future of AI: Reshaping Society

Quick Answer: Why Is Empathy Critical for Postdigital Government?

Government Insight: U.S. Federal AI Executive Order Opportunities and Risks

Seize the Opportunities: How Can Will U.S. Policy on AI Impact Business Strategy?

AI Agents
Analysis By: Tom Coshow, Haritha Khandabattu

Benefit Rating: High

Market Penetration: 5% to 20% of target audience

Maturity: Adolescent

Gartner, Inc. | G00828523 Page 64 of 108

This research note is restricted to the personal use of [email protected].

 Definition:

AI agents are autonomous or semiautonomous software entities that use AI techniques to

perceive, make decisions, take actions and achieve goals in their digital or physical
environments.

Why This Is Important

AI agents have the ability to make decisions and take action in their target environment to
achieve organizational goals. Using AI practices and techniques such as LLMs,
organizations are creating and deploying AI agents to achieve complex tasks.

Business Impact

AI agents have the potential to:

■ Revolutionize a broad range of industries and environments with their ability to

automate tasks from consumer, industrial, data analytics, content creation and
logistics.

■ Make informed decisions and interact intelligently with their surroundings.

Drivers

■ Generative AI breakthroughs: Reasoning models and LAMs advance the ability to

plan a complex series of actions.

■ Multimodal understanding: The ability to use diverse modalities like vision, audio
and language enables more general and flexible AI agents. This allows automatic
adaptation to changes in the workflow, user interface or API. With this, one can
create advanced workflows without explicit programming, significantly reducing the
development time and effort for automation.

■ Increased decision-making complexity: AI is increasingly used in real-world

engineering problems containing complex systems, where large networks of
interacting parts exhibit emergent behavior that cannot be easily predicted. AI agents
can learn, plan and execute in complex environments.

■ Composite AI, including neurosymbolic models: Advances in models that improve

planning and problem solving are enabling more complex AI agents. AI agents can
utilize a wide variety of AI practices to forecast, make decisions and plan.

Gartner, Inc. | G00828523 Page 65 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles
■ Vulnerabilities: Due to the complexity of the AI agent system, all components face
various potential vulnerabilities such as access security, data security and
governance.

■ Lack of trust: Users are unsure whether they can trust the technology to accurately
predict and execute tasks independently. Without a human in the loop, agents may
take multiple consequential actions in rapid succession and bring about significant
impacts before a human notices.

■ Interpretability and oversight: Action policies may be opaque and have poor
explainability, requiring mechanisms for human interpretability, oversight and
control.

■ Pace of change: The technologies deployed, from models to tooling, and the
framework options available are changing rapidly, making it difficult for
organizations to define their roadmap.

User Recommendations

■ Incorporate AI agents into strategic planning by investing in understanding their

capabilities and potential applications in various environments, considering their
increasing autonomy and wide-ranging usability.

■ Investigate the possibilities of utilizing multiagent systems, collectives of AI agents,

that can operate both collaboratively and independently, enhancing adaptability and
flexibility in response to different tasks and scenarios.

■ Promote the development and integration of the use of a variety of AI practices,

enabling learning, negotiation and decision-making capabilities.

Sample Vendors

Amazon Web Services; Autogen; CrewAI; Google; LangChain; Maisa; OneReach.ai;

Salesforce; UnifyApps; Zhipu AI

Gartner Recommended Reading

Innovation Insight: AI Agents

Build AI Agent Services to Revolutionize Client Operations

Innovation Insight for the AI Agent Platform Landscape

Gartner, Inc. | G00828523 Page 66 of 108

This research note is restricted to the personal use of [email protected].

 AI Agents: Are You Ready to Set Your AI Free?

Emerging Patterns for Building LLM-Based AI Agents

AI-Ready Data
Analysis By: Roxane Edjlali, Mark Beyer, Svetlana Sicular, Ehtisham Zaidi

Benefit Rating: Transformational

Market Penetration: 1% to 5% of target audience

Maturity: Adolescent

Definition:

AI-ready data is determined through the data’s ability to prove its fitness for use for
specific AI use cases. Proof of readiness comes from the assessment of its
representativeness evaluated by its alignment to the use case, support for continuous
data qualification, and ensuring data and AI governance. As a result, AI-ready data can
only be determined contextually to the AI use case and the AI technique used, which
forces new approaches to data management.

Why This Is Important

With the rise of pretrained off-the-shelf models and hype from generative AI (GenAI), data
management leaders are at the forefront of creating data strategies for AI. Chief data and
analytics officers and data management leaders must quickly respond to rising AI-ready
data demands by delivering AI-ready data to support AI use cases. Organizations not
investing in AI-ready data practices can increasingly fail to deliver to business objectives
and face data and AI governance issues that can lead to erroneous results and financial
risk.

Business Impact

Organizations that invest in AI at scale need to evolve their data management practices
and capabilities not only to preserve the evergreen classical ideas of data management
but also to extend them to AI. It will be critical to provision AI-ready data iteratively to cater
to existing and upcoming business demands, ensure trust, avoid risk and compliance
issues, preserve intellectual property, and reduce bias and hallucinations.

Gartner, Inc. | G00828523 Page 67 of 108

This research note is restricted to the personal use of [email protected].

 Drivers
■ Data is becoming the main source of differentiation and value from these pretrained
models. Models, especially for GenAI, increasingly come from vendors rather than
delivered in-house.

■ According to the 2024 Gartner Evolution of Data Management Survey, 57% of

organizations estimate their data is not AI-ready, and among the remaining 43% that
do, the readiness assessment demonstrates gaps.

■ According to the 2024 Gartner AI Mandates for the Enterprise Survey, participants
report that data availability or quality is the No. 1 barrier to successful AI
implementation.

■ Rapid progress of AI poses new challenges in organizing and managing AI data. A

cycle of augmented data management techniques better suited for meeting AI data
requirements is expected. Data ecosystems on the foundation of data fabric
architecture indicate the beginning of this new cycle.

■ Augmented data management capabilities and tools greatly benefit from AI. AI
techniques offer new data-centric approaches, such as automated feature
engineering or assisted data engineering and code generation using retrieval-
augmented generation.

■ GenAI is removing the distinction between structured and unstructured data, thereby
requiring data management to adapt to new uses.

Gartner, Inc. | G00828523 Page 68 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ The AI community remains mostly unaware of data management capabilities,

practices and tools that can greatly benefit AI development and deployment. The
lack of information can lead to challenges when scaling prototypes in production.
Traditional data management also ignores AI-specific considerations such as data
bias, labeling and drift; this is changing but slowly.

■ Responsible AI requires new governance approaches for both the data and AI model.
These AI-specific data practices are not yet part of traditional data governance in
most enterprises.

■ Assuming AI models have addressed all potential data management issues once
deployed is a fallacy. Deployment considerations such as ongoing drift monitoring
require ongoing data management activities and practices.

■ AI developers are focused on the use case context as opposed to independent

validation and reuse, affecting effective production use and reusability across use
cases.

Gartner, Inc. | G00828523 Page 69 of 108

This research note is restricted to the personal use of [email protected].

 User Recommendations
■ Formalize AI-ready data as a dedicated practice as part of your overall data
management strategy. Implement active metadata management, data quality,
observability, integration and fabric as foundational components of this strategy.

■ Train data engineers in support of AI and train AI specialists in data management.

■ Support AI model development in a data-centric way due to the dependency of AI

models on representative data. Diversify data, models and people to ensure AI value
and avoid involuntary bias.

■ Utilize data management expertise, AI engineering, DataOps and MLOps approaches

to support the AI life cycle. Include data management requirements when deploying
AI models.

■ Develop data monitoring and data governance metrics to ensure that your AI models
produce the correct output continuously.

■ Define and measure minimum data standards for AI readiness of data early for each
use case and continuously prove data fitness when taking AI to scale. These include
checking lineage, quality and governance assessment, versioning and automated
testing.

■ Investigate data management tools rich in augmented data management

capabilities that can integrate well with AI tools that have created disruptive data-
centric AI capabilities, like multimodal data fabric.

Gartner Recommended Reading

A Journey Guide to Delivering AI Success Through ‘AI -Ready’ Data

Quick Answer: What Makes Data AI-Ready?

Innovation Insight: How Generative AI Is Transforming Data Management Solutions

Become AI-Ready by Focusing on Foundational Data Tools and Technologies

How to Evaluate AI Data Readiness

AI Engineering
Analysis By: Soyeb Barot, Cuneyd Kaya, Leinar Ramos, Anthony Mullen

Gartner, Inc. | G00828523 Page 70 of 108

This research note is restricted to the personal use of [email protected].

 Benefit Rating: Transformational

Market Penetration: 5% to 20% of target audience

Maturity: Early mainstream

Definition:

AI engineering is the foundational discipline for enterprise delivery of AI and generative AI

(GenAI) solutions at scale. The discipline unifies DataOps, ModelOps and DevOps
pipelines to create a coherent development, deployment (hybrid, multicloud, edge), and
operationalization framework for AI-based systems.

Why This Is Important

The demand for AI solutions has dramatically increased, driven by the unrelenting
adoption of GenAI techniques. Few organizations have built the data, AI model
management and DevOps foundations required to move pilot projects to production,
much less operate portfolios of AI solutions at scale. In addition, an appropriate operating
model is required that enables cross-collaboration across teams with diverse skills to
build AI systems. Hence, to meet the demands for scaling AI solutions, enterprises must
establish consistent pipelines supporting the development, deployment, reuse, governance
and maintenance of AI models (statistical, machine learning, generative, deep learning,
graph, linguistic and rule-based).

Business Impact

AI engineering enables organizations to establish and grow a high-value portfolio of AI

solutions consistently and securely. Almost every AI solution is built using an ensemble of
models, using multiple AI techniques, thereby resulting in composite AI systems. There are
significant engineering, process and culture challenges to address as part of building and
deploying composite AI systems at scale. With defined AI engineering approaches —
DataOps, ModelOps and DevOps — it is possible to deploy models into production in a
structured, repeatable factory model. AI engineering also enables establishing process
flow to ensure the AI engineering team works closely with business stakeholders
throughout the development process of AI-based solutions.

Gartner, Inc. | G00828523 Page 71 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ The elimination of traditional siloed approaches to data management and AI

engineering doubles the data engineering effort and reduces impedance mismatches
across data ingestion, processing, model engineering and deployment, which
inevitably drift further once the AI models are in production.

■ DataOps, ModelOps (including LLMOps) and DevOps provide best practices for
moving artifacts through the AI development life cycle. Standardization across data
and model pipelines accelerates the delivery of AI solutions irrespective of the
approach — retrieval-augmented generation (RAG) or fine-tuning techniques
alongside implementations with models built using diverse AI techniques.

■ AI engineering enables discoverable, composable and reusable data and AI artifacts

(such as data catalogs, knowledge graphs, code repositories, reference architectures,
feature stores and model stores) across the enterprise technical architecture. These
are essential for scaling AI enterprisewide.

■ AI engineering practices, processes and tools must be adapted to address GenAI.

GenAI-specific adaptations include support for prompt engineering, vector
DBs/knowledge graphs, architecting and deploying multiagent systems, and
interactive deployment models.

■ AI engineering tools can be subdivided into model-centric and data-centric tools.

Terms such as DataOps, LLMOps, or broader terms such as ModelOps or MLOps, are
used frequently. However, we believe they are all a subset of AI engineering that are
DevOps best practices implemented to operationalize specific portions of the AI
development life cycle.

■ Finally, AI engineering makes it possible to orchestrate solutions across hybrid,

multicloud, edge AI or Internet of Things (IoT).

Gartner, Inc. | G00828523 Page 72 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles
■ The rapid expansion of the Ops family has led to an influx of newer, yet marginally
nuanced, Ops terms that are overwhelming AI leaders. Because of this ambiguity, AI
leaders often fail to prioritize and pursue the right Ops disciplines, leading to an
inferior AI operational state.

■ AI engineering requires simultaneous development of pipelines across domains

alongside maturity across the platform infrastructure.

■ It requires integrating full-featured solutions with specific tools, enabling

operationalization capabilities, to address enterprise architecture gaps with minimal
functional overlap. These include gaps around extraction, transformation and
loading stores, feature stores, model stores, model monitoring, pipeline observability,
and governance.

■ AI engineering requires cloud maturity and possible rearchitecting, or the ability to

integrate data and AI model pipelines across deployment contexts. Potential
complexity and management of analytical and AI workloads alongside costs may
deter organizations that are in the initial phases of AI initiatives.

■ Enterprises often seek “unicorn” experts to productize AI platforms and solutions.

Spot-fix vendor solutions will bloat costs and potentially complicate already intricate
integration and model management tasks.

Gartner, Inc. | G00828523 Page 73 of 108

This research note is restricted to the personal use of [email protected].

 User Recommendations

■ Maximize business value from ongoing AI initiatives by establishing an AI

engineering practice that streamlines data, models and implementation pipelines.

■ Simplify data and analytics pipelines by identifying the capabilities required to

operationalize end-to-end AI development platforms and build AI-specific toolchains.

■ Use point solutions sparingly and only to plug feature/capability gaps in fully
featured DataOps, MLOps, ModelOps and tools. Open-source technologies such as
Airflow and Kafka, play an important role knitting together a comprehensive solution.

■ Leverage cloud service provider environments as foundational to build AI

engineering. At the same time, rationalize your data, analytics and AI portfolios as
you migrate to the cloud.

■ Adopt a platform approach with GenAI by investing in centralized AI engineering

tools for automation, governance and use-case enablement across a broad set of AI
models and cloud service providers.

■ Adopt software delivery best practices by establishing a common understanding of

the various aspects of your solution, avoiding silos and engaging with business
stakeholders.

■ Upskill data engineering and platform engineering teams to adopt tools and
processes that drive continuous integration/continuous development for AI artifacts.

Sample Vendors

Amazon Web Services; Anyscale; CoreWeave (Weights & Biases); Dataiku; DataRobot;
Domino Data Lab; Google; Microsoft; NVIDIA (OctoAI); Unstructured

Gartner Recommended Reading

Top Strategic Technology Trends for 2022: AI Engineering

Demystify the Ops Landscape to Scale AI Initiatives: A Gartner Trend Insight Report

Case Study: AI Model Operations at Scale (Fidelity)

AI Engineering Best Practices for Data Scientists

Gartner, Inc. | G00828523 Page 74 of 108

This research note is restricted to the personal use of [email protected].

 Responsible AI
Analysis By: Svetlana Sicular, Philip Walsh

Benefit Rating: Transformational

Market Penetration: 5% to 20% of target audience

Maturity: Early mainstream

Definition:

Responsible artificial intelligence (RAI) is an umbrella term for aspects of making

appropriate business and ethical choices when adopting AI. These include business and
societal value, risk, trust, transparency, fairness, bias mitigation, explainability,
sustainability, accountability, safety, privacy and regulatory compliance. RAI encompasses
organizational responsibilities and practices that ensure positive, accountable and ethical
AI development and operation.

Why This Is Important

Wide adoption of AI resulted in separation of RAI mostly into individual focus areas, AI
governance, and trust, risk and security management (TRiSM). Furthermore, jurisdictions
and industry regulations concretize practices that were less defined under the RAI
umbrella. While the term responsible AI is still in use, enterprises will continue
emphasizing specific areas and focus on their nuanced AI goals, such as risk, privacy,
compliance, ethics, AI applications evaluation, and ensuring AI-ready data.

Business Impact

Responsible attitudes toward AI are required from every role in the organization. RAI
assumes accountability for AI development and use at the individual, organizational and
societal levels. If AI governance and TRiSM are practiced by designated groups, RAI
extends its reach to all stakeholders involved in the AI process. Concrete RAI practices,
such as preserving privacy and debiasing AI, protect organizations by ensuring AI
technologies are beneficial, safe, ethical and trustworthy.

Drivers

RAI helps AI participants develop, implement, utilize and address the various drivers they
face. With widening AI adoption, the RAI drivers are becoming more important and are
better understood by vendors, buyers, society and legislators:

Gartner, Inc. | G00828523 Page 75 of 108

This research note is restricted to the personal use of [email protected].

 ■ The adoption of generative AI (GenAI) raises new concerns, such as hallucinations,
leaked sensitive data, copyright issues and reputational risks, that bring new actors
in RAI (for example, in security, legal and procurement).

■ Leading vendors are offering indemnification of their GenAI offerings, making

customers more confident as part of their RAI approaches: although a good step,
these are still incomplete.

■ The organizational driver of RAI assumes the need to strike a balance between the
business value and associated risks within regulatory, business and ethical
boundaries. This includes considerations such as reskilling employees to adapt to AI
technologies and safeguarding intellectual property.

■ The societal driver includes resolving AI safety for societal well-being versus limiting
human freedoms. Existing and pending legal guidelines and regulations, such as the
EU’s Artificial Intelligence Act, make RAI a necessity.

■ The customer/citizen driver is based on fairness and ethics and requires reconciling
privacy with convenience. Customers/citizens may be willing to share their data in
exchange for certain benefits.

■ AI affects all ways of life and touches all societal strata; hence, the RAI challenges
are multifaceted and cannot be easily generalized, therefore, organizations address
concrete items under the RAI umbrella. New problems will continue to arise with
rapidly evolving technologies and their uses.

Gartner, Inc. | G00828523 Page 76 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ RAI may look good on paper, but poorly defined accountability for RAI renders it
ineffective in reality.

■ Organizations lack awareness of AI’s unintended consequences. Many turn to RAI

practices only after they experience AI’s negative effects, whereas prevention is
simpler.

■ Most AI regulations are still in draft. AI products’ adoption of regulations for privacy
and intellectual property makes it challenging for organizations to ensure
compliance and avoid all possible liability risks.

■ Rapidly evolving AI technologies, including tools for explainability, reasoning, bias

and hallucinations detection, privacy protection and some regulatory compliance, lull
organizations into a false sense of responsibility, while mere technology is not
enough. A disciplined AI risk, ethics and governance approach is necessary, in
addition to technology.

■ Creating RAI principles and operationalizing them without regularly measuring the
progress makes it hard to sustain RAI practices.

User Recommendations

■ Identify and prioritize RAI focus areas for your AI strategy. Publicize consistent
approaches across all RAI focus areas. Typical areas of RAI in the enterprise are
fairness, bias mitigation, ethics, risk management, security, privacy, reliability,
sustainability and regulatory compliance.

■ Designate a champion for each use case who will be accountable for the responsible
development and use of AI.

■ Define the AI life cycle framework. Address RAI in all phases of this cycle. Address
hard trade-off questions.

■ Provide training for applicable RAI focus areas to personnel. Include AI literacy and
critical thinking as part of the training.

■ Operationalize RAI principles. Ensure diversity of participants and enable them to

easily voice AI concerns.

■ Participate in industry or societal AI groups. Learn best practices and contribute your
own because everybody will benefit from this exchange. Ensure policies account for
the needs of any internal or external stakeholders.

Gartner, Inc. | G00828523 Page 77 of 108

This research note is restricted to the personal use of [email protected].

 Gartner Recommended Reading

Trends in Action: Employee Activism Drives Adoption and Norms for Responsible AI

Create an AI Literacy Roadmap to Drive Responsible and Productive AI

Critical Insights: Impact of U.S. Federal Policy Changes on Responsible AI

ModelOps
Analysis By: Cuneyd Kaya, Soyeb Barot

Benefit Rating: Transformational

Market Penetration: 5% to 20% of target audience

Maturity: Adolescent

Definition:

Model operationalization (ModelOps) is primarily focused on the end-to-end governance

and life cycle management of advanced analytics, AI and decision models, such as
models based on machine learning (ML), generative AI (GenAI) and knowledge graphs.

Why This Is Important

ModelOps helps companies in standardizing, scaling and augmenting their analytics, AI

and GenAI initiatives. It helps organizations to move their models from the lab
environments into production. MLOps primarily focuses on monitoring, operationalization
and governance of ML models, while ModelOps assists with the same for all advanced
analytics, decision and AI models, including GenAI and retrieval-augmented generation
(RAG) systems.

Business Impact

ModelOps as a practice:

■ Provides the capability for the management and operationalization of diverse AI

(including generative AI), analytics and decision intelligence systems

■ Enables the complex subsystems required for such systems with observability,
versioning, monitoring, data and model orchestration, guardrails, experimentation
and explainability.

Gartner, Inc. | G00828523 Page 78 of 108

This research note is restricted to the personal use of [email protected].

 ■ Ensures collaboration among a wider business, development and deployment
community and the ability to associate model outcomes with business KPIs

■ Is an important tech capability part of AI engineering; ModelOps’ best practices make

it possible to deploy models into production in a structured, repeatable factory model

Drivers

■ Modern AI systems are being built with a symbiotic combination of generative and
classic AI models, agents and intelligent software capabilities. As the number of
advanced analytics, AI and decision models at organizations increases,
organizations will have to manage different types of prepackaged or custom-made
models in production. In doing so, it becomes crucial for organizations to ensure
these models are working as intended, maintaining accuracy and reliability across
various applications.

■ Organizations want to be more agile and responsive to changes within their

advanced analytics and AI pipelines not only with models, but also with data,
application and infrastructure.

■ ModelOps provides a framework to separate responsibilities across various teams

for how models (including GenAI, foundational models, analytics, ML, physical,
simulation and symbolic) are built, tested, deployed and monitored across different
environments (for example, development, test and production). This enables better
productivity and collaboration and it lowers failure rates.

■ ModelOps provides tools to address model degradation via drift and bias. In other
scenarios, enabling model runtime enforcements, explainability and integrity is
paramount.

■ Organizations don’t want to deploy an unlimited number of open-source offerings to

manage ModelOps, but there are few comprehensive solutions that provide end-to-
end capabilities in every domain of model operationalization. Moreover, not every
capability is required immediately. Often, versioning, monitoring and model
orchestration precede the full implementation of feature stores, pipelines and
observability.

■ GenAI will require an increased focus on testing and the introduction of capabilities
to version, manage and automate prompts, routers and retrieval-augmented
generation systems. Fine-tuning will also require enhanced ModelOps capabilities to
manage transformations for complex domain and function training datasets.

Gartner, Inc. | G00828523 Page 79 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Organizations using different types of models often don’t build the right Ops, model
usage policy and management capabilities until they already have a chaotic
landscape of unmanaged advanced analytic and AI systems.

■ Not all analytical techniques currently benefit from mature operationalization

methods. Because the spotlight has been on ML techniques, MLOps benefits from a
more evolved AI practice, but some models, like agentic modeling and optimization
techniques, require more attention in ModelOps practices and platforms.

■ Organizations are struggling to get GenAI into production, due to their data not being
consumable by AI models, along with data privacy, security and regulatory concerns.

■ Organizations may adopt ModelOps platform capabilities that they don’t

immediately need. At the same time, organizations that are siloed and fail to adopt a
comprehensive ModelOps strategy create redundancy in effort with respect to
operationalization.

User Recommendations
■ Buy ModelOps capabilities integrated into your primary AI platforms. Enrich these
capabilities with best-of-breed open-source or proprietary ModelOps offerings where
unique problems, like feature stores or observability, require enhanced solutions.

■ Utilize ModelOps best practices across composite AI, data, models and applications
to ensure transition, reduce friction and increase value generation.

■ Recruit/upskill additional engineers who can master ModelOps on AI systems that

utilize unstructured data, search and retrieval, graph and optimization.

■ Encourage collaboration between data science and development and deployment

teams; empower teams to make decisions to automate, scale and bring stability to
the analytics and AI pipeline.

■ Collaborate with software engineering teams to scale ModelOps. Offloading

operationalization responsibilities to production support teams enables increased
ModelOps specialization and sophistication across the ecosystem of complex AI-
enabled applications.

Sample Vendors

Amazon Web Services; CoreWeave (Weights & Biases); Databricks; Dataiku; DataRobot;
Domino Data Labs; Google Cloud; IBM; Microsoft Azure; ModelOp

Gartner, Inc. | G00828523 Page 80 of 108

This research note is restricted to the personal use of [email protected].

 Gartner Recommended Reading

Demystify the Ops Landscape to Scale AI Initiatives: A Gartner Trend Insight Report

Launch an Effective Machine Learning Monitoring System

Toolkit: Delivery Metrics for DataOps, Self-Service Analytics, ModelOps and MLOps

Reference Architecture Brief: MLOps Architecture

Critical Capabilities for Data Science and Machine Learning Platforms, AI Engineering

Gartner, Inc. | G00828523 Page 81 of 108

This research note is restricted to the personal use of [email protected].

 Sliding into the Trough
Foundation Models
Analysis By: Arun Chandrasekaran

Benefit Rating: Transformational

Market Penetration: More than 50% of target audience

Maturity: Mature mainstream

Definition:

Foundation models are large-parameter models that are trained on a broad gamut of
datasets in a self-supervised manner. They are mostly based on transformer or diffusion
deep neural network architectures and are increasingly becoming multimodal. They are
called foundation models because of their critical importance and applicability to a wide
variety of downstream use cases. This broad applicability is due to the pretraining and
versatility of the models.

Why This Is Important

Foundation models are an important step forward for AI due to their massive pretraining
and wide use-case applicability. They can deliver state-of-the-art capabilities with higher
efficacy than their predecessors. They’ve become the go-to architecture for natural
language processing and have also been applied to computer vision, audio and video
processing, and software engineering use cases.

Business Impact

With their potential to enhance applications across a broad range of enterprise use cases,
foundation models are having a wide impact across vertical industries and business
functions. Their impact has accelerated, with a growing ecosystem of startups building
enterprise applications on top of them. Foundation models will advance digital
transformation within the enterprise by improving workforce productivity, automating and
enhancing customer experience, and enabling rapid, cost-effective creation of new
products and services.

Drivers

Gartner, Inc. | G00828523 Page 82 of 108

This research note is restricted to the personal use of [email protected].

 ■ Quicker time to value: Foundation models can effectively deliver value through
prebuilt APIs, prompt engineering, retrieval-augmented generation or further fine-
tuning. While fine-tuning may enable more customization, the other three options are
less complex, quicker and cheaper.

■ Superior performance across multiple domains: The difference between foundation

models and prior neural network solutions is stark. The large pretrained models can
produce coherent text, code, images, speech and video at a scale and accuracy not
possible before and are increasingly becoming better at reasoning tasks.

■ Fast-paced innovation: The past year has seen an influx of foundation models,
along with smaller, pretrained domain-specific models built from them. Most of
these are available as cloud APIs or open-source projects, further reducing the time
and cost to experiment and driving quicker enterprise adoption.

■ Productivity gains: Foundation models are having an impact across broad swaths
of enterprise business functions as their ability to automate tasks gets wider.
Business functions such as marketing, customer service and IT (especially software
engineering) are areas where clients are seeking initial gains.

Obstacles
■ Flawed results: Although a significant advance, foundation models still require
careful training and guardrails. Because of their training methods and black-box
nature, they can deliver unacceptable results or hallucinations. They also can
propagate downstream any bias or copyright issues in the datasets.

■ Require appropriate skills and talent: As with all AI solutions, the end result depends
on the skills, knowledge and talent of the trainers and users, particularly for prompt
engineering and fine-tuning.

■ Data integration: Adapting foundation models to specific enterprise use cases often
requires building RAG pipelines, fine-tuning or significant prompt engineering
automation. While the know-how of how to implement this is growing, highly
effective techniques to ground these models remain both technically complex and
expensive.

User Recommendations

Gartner, Inc. | G00828523 Page 83 of 108

This research note is restricted to the personal use of [email protected].

 ■ Plan to introduce foundation models into existing speech, text or coding domains.
If you have any older language processing systems, moving to a transformer-based
model could significantly improve performance. Knowledge search, summarization
and content generation are popular emerging use cases across industries.

■ Start with models that have superior ecosystem support and adequate enterprise
guardrails around security and privacy and are more widely deployed.

■ Be objective about the adequate balance between accuracy, costs, security and
privacy, and time to value when selecting foundation models to determine the
appropriate model needed. Beware of building models from scratch, given the
complexity and steep costs.

■ Educate developers and data and analytics teams on prompt engineering and other
advanced techniques needed to steer these models.

■ Designate an incubation team to monitor industry developments, communicate the

“art of the possible,” experiment with business units and share valuable lessons
learned companywide.

Sample Vendors

Alibaba Group; Anthropic; Cohere; DeepSeek AI; Google; IBM; Meta; Microsoft; Mistral AI;
OpenAI

Gartner Recommended Reading

Innovation Guide for Generative AI Models

Answering IT Leaders’ Top 10 Questions on Open Generative AI Models

Explore Small Language Models for Specific AI Scenarios

Synthetic Data
Analysis By: Arun Chandrasekaran, Alys Woodward, Anthony Mullen

Benefit Rating: High

Market Penetration: 5% to 20% of target audience

Maturity: Early mainstream

Gartner, Inc. | G00828523 Page 84 of 108

This research note is restricted to the personal use of [email protected].

 Definition:

Synthetic data is a class of data that is artificially generated rather than obtained from
direct observations of the real world. Synthetic data is used as a proxy for real data in a
wide variety of use cases, including data anonymization, AI and machine learning (ML)
development, data sharing, and data monetization.

Why This Is Important

Obtaining and labeling real-world data for AI development is a time-consuming and

expensive task. Specifically for use cases such as training models for autonomous
vehicles, collecting real data for 100% coverage of edge cases is either practically
impossible or incredibly hard. These challenges can be remedied with synthetic data. It
can be generated quickly, cost-effectively, and without personally identifiable information
(PII) or protected health information (PHI), making it a valuable technology for privacy
preservation. The rise of frontier AI models has highlighted synthetic data as a cost-
effective means to build scalable models.

Business Impact

Adoption is increasing across various industries, and Gartner predicts a further rise as
synthetic data:

■ Avoids using PII when training AI models via synthetic variations of original data or
synthetic replacement of parts of data.

■ Reduces cost and saves time in ML development.

■ Improves AI performance as more “fit for purpose” training data leads to better
outcomes.

■ Enables organizations to pursue new use cases for which minimal real data is
available.

■ Addresses fairness issues, such as bias and toxicity, more efficiently.

■ Enables software testing on realistic yet private data, without legislation risks or in
the absence of such data.

Gartner, Inc. | G00828523 Page 85 of 108

This research note is restricted to the personal use of [email protected].

 Drivers
■ In regulated industries such as healthcare and finance, buyer interest is growing, as
synthetic tabular data can be used to preserve privacy in AI training data.

■ To meet the increasing demand for synthetic data for natural language automation
training, especially for chatbots and speech applications, vendors are bringing new
offerings to market, often to train domain generative AI (GenAI) models. This is
expanding the vendor landscape and driving synthetic data adoption.

■ Synthetic data applications have expanded beyond automotive and computer vision
use cases to include data monetization, external analytics support, platform
evaluation and the development of test data.

■ Transformer and diffusion architectures, the architectural foundations for GenAI, are
enabling synthetic data generation at quality and precision levels not seen before. AI
simulation techniques are improving synthetic data quality by better recreating real-
world representations.

■ There is scope for expansion to other data types. While tabular, image, video, text
and speech applications are common, R&D labs are expanding the concept of
synthetic data to graphs and multimodal AI. Synthetically generated graphs will
resemble but not overlap the original. As organizations begin to use graph
technology more, we expect this method to mature and drive adoption.

■ As data providers for training frontier AI models raise their data access costs,
synthetic data is gaining traction as an economic alternative.

Gartner, Inc. | G00828523 Page 86 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles
■ Synthetic data can have implicit bias problems, miss natural anomalies, be
complicated to develop, or not contribute any new information to existing, real-world
data.

■ While synthetic data reduces privacy risks, some industries (e.g., healthcare, finance)
still face regulatory uncertainty about whether it can fully replace or augment real
data for compliance purposes.

■ Synthetic data generation methodologies lack standardization.

■ Validating the accuracy of synthetic data is difficult. It can be challenging to know

for sure whether a synthetic dataset accurately captures the underlying real-world
environment.

■ Buyers are still confused about when and how to use the technology due to the lack
of skills.

■ Enterprises have legacy data pipelines, and integrating synthetic data into existing
data lakes, analytics systems and AI workflows often requires additional effort,
tooling and infrastructure.

■ There may be a level of user skepticism as data may be perceived to be “inferior” or

“fake.”

User Recommendations
■ Identify areas in your organization where data is missing, incomplete or expensive to
obtain, and is thus, currently blocking AI initiatives.

■ Establish clear policies on synthetic data life cycle management, storage and access
control.

■ Work with legal and compliance teams to ensure synthetic data aligns with industry
regulations, such as General Data Protection Regulation, Health Insurance Portability
and Accountability Act and the California Consumer Privacy Act.

■ Educate internal stakeholders through training programs on the benefits and

limitations of synthetic data. Institute guardrails to mitigate challenges such as user
skepticism and inadequate data validation.

■ Measure and communicate the business value and the success and failure stories of
synthetic data initiatives.

Gartner, Inc. | G00828523 Page 87 of 108

This research note is restricted to the personal use of [email protected].

 Sample Vendors

Anyverse; Betterdata; Bifrost; Gretel; MOSTLY AI; Parallel Domain; Rendered.ai; SAS;
Tonic.ai; YData

Gartner Recommended Reading

Innovation Guide for Generative AI Models

Executive Briefing on Emerging Technology: Synthetic Data

Market Guide for Data Masking and Synthetic Data

Edge AI
Analysis By: Eric Goodness

Benefit Rating: High

Market Penetration: 20% to 50% of target audience

Maturity: Adolescent

Definition:

Edge AI is the use of AI techniques embedded in non-IT products (consumer/commercial,

industrial), Internet of Things (IoT) endpoints, gateways and edge servers. Capabilities
span consumer, commercial and industrial uses, such as mobile devices, autonomous
vehicles, enhanced medical diagnostics and streaming analytics. While predominantly
focused on AI inference, more sophisticated systems include local training capabilities to
optimize models at the edge.

Why This Is Important

Many edge computing use cases are latency-sensitive and data-intensive, and require a
level of autonomy and data sovereignty, for local decision intelligence. Such needs have
resulted in AI deployment in a wide range of edge computing solutions. Edge AI allows
industries, in hazardous and/or highly regulated environments, to apply various AI and
machine learning (ML) techniques in operational environments. These applications
include distributed, resource-constrained assets, an ability to benefit from improved
decision support providing a close-feedback loop, with automated machine tasks, that
enables asset reliability.

Gartner, Inc. | G00828523 Page 88 of 108

This research note is restricted to the personal use of [email protected].

 Business Impact
■ Real-time data analysis and decision intelligence.

■ Improved operational efficiency, such as visual inspection systems for quality

management, output and process efficiency.

■ Enhanced customer experience (CX) from AI feedback embedded within products.

■ Reduced connectivity costs with fewer data journeys between the edge and cloud.

■ Persistent functionality, independent of connectivity.

■ Reduced storage demand as only prioritized data is passed to core systems.

■ Preserved data privacy at the endpoint.

Drivers

Overall, edge AI has benefited from improvements in the capabilities of AI, including:

■ The maturation of MLOps and ModelOps tools and processes that support ease of
use across a broader set of features spanning the wider MLOps functions. Initially,
many companies came to market with a narrowcast focus on model compression.

■ The improved performance of combined ML techniques and an associated increase

in data availability (such as time-series data from industrial assets).

Business demand for new and improved outcomes, solely achievable from the use of AI at
the edge, include:

■ Reducing full-time equivalents with vision-based solutions used for surveillance or

inspections.

■ Improving manufacturing production quality by automating various processes.

■ Optimizing operational processes across industries.

■ New approaches to CX, such as personalization on mobile devices or changes in

retail from edge-based smart check-out points of sale.

■ Privacy-preserving edges.

Additional drivers include:

Gartner, Inc. | G00828523 Page 89 of 108

This research note is restricted to the personal use of [email protected].

 ■ An increasing number of users are upgrading legacy systems and infrastructure in
“brownfield” environments. By using MLOps platforms, AI software can be hosted
within an edge computer or a gateway (aggregation point) or embedded within a
product with the requisite compute resources.

■ More manufacturers are embedding AI in the endpoint as an element of product

servitization. In this architecture, IoT endpoints, such as in automobiles, home
appliances and commercial building infrastructure, are capable of running AI models
to interpret data captured by the endpoint and drive some of the endpoints’
functions.

■ Rising demand for R&D in training has decentralized AI models at the edge for
adaptive AI. These emerging solutions are driven by explicit needs such as privacy
preservation or the requirement for machines and processes to run in disconnected
(from the cloud) scenarios.

Obstacles

■ Edge AI is constrained by the limitations of the equipment deployed, such as form

factor, power budget, data volume, decision latency and security.

■ Systems deploying AI techniques can be nondeterministic. This will impact edge AI

applicability in certain use cases, especially where safety and security requirements
are primary.

■ The autonomy of edge AI-enabled solutions, built on some ML and deep learning
techniques, often presents questions of trust, especially where the inferences are not
readily explainable. As adaptive AI solutions increase, these issues will increase if
identical models deployed to equivalent endpoints begin to evolve diverging
behaviors.

■ The lack of quality and sufficient data for training is a universal challenge.

■ Deep learning in neural networks is a compute-intensive task, often requiring the use
of high-performance chips with corresponding high-power budgets. This limits
deployment locations where small-form factors and low-power requirements are
paramount.

Gartner, Inc. | G00828523 Page 90 of 108

This research note is restricted to the personal use of [email protected].

 User Recommendations

■ Determine whether the use of edge AI provides suitable cost-benefit improvements or

whether traditional centralized data analytics and AI methodologies are adequate
and scalable.

■ Evaluate when to consider AI at the edge versus a centralized solution. Good

candidates for edge AI are applications that have high communications costs, are
sensitive to latency, require real-time responses or ingest high volumes of data at the
edge.

■ Assess the different technologies available to support edge AI and the viability of the
vendors offering them. Many potential vendors are startups that may have
interesting products but limited support capabilities.

■ Use edge gateways and servers as the aggregation and filtering points to perform
most of the edge AI and analytics functions. Make an exception for compute-
intensive endpoints, where AI-based analytics can be performed on the devices
themselves.

Sample Vendors

Edge Impulse; IFS (Falkonry); Johnson Controls; Pratexo; Synadia Communications

Gartner Recommended Reading

Emerging Tech Impact Radar: Edge Artificial Intelligence

Innovation Insight for Edge AI

Emerging Tech: Differentiate With an Edge AI Benchmarking Strategy

Market Guide for Edge Computing

Emerging Tech: Empower Outcome-Centric IoT With AI

Generative AI
Analysis By: Svetlana Sicular

Benefit Rating: Transformational

Market Penetration: More than 50% of target audience

Gartner, Inc. | G00828523 Page 91 of 108

This research note is restricted to the personal use of [email protected].

 Maturity: Adolescent

Definition:

Generative AI (GenAI) technologies can generate new derived versions of content,

strategies, designs and methods by learning from large repositories of original source
content. Generative AI has profound business impacts, including on content discovery,
creation, authenticity and regulations; automation of human work; and customer and
employee experiences.

Why This Is Important

GenAI is becoming real in enterprises. AI leaders from the 2024 Gartner AI Mandates for
the Enterprise Survey reported an average spend of $1.9 million in fiscal year 2024 on
GenAI initiatives, which reflects a belief in further GenAI potential. Governments are
committing large funds to GenAI; vendors continue fast innovation, advancing model
performance, multimodality, reasoning and agentic capabilities. Research of training data,
explainability, fine-tuning, distillation and other aspects of GenAI exploitation is fast-paced
and is reflected in commercial and open-source solutions.

Business Impact

GenAI has a strong momentum for expansion and deeper integration into business
workflows across various business functions and industries. Fully integrated tools,
accompanied by AI governance practices, robust education and IT support, enable
enterprises to tackle critical business processes. Multimodal GenAI opens new
opportunities in life sciences, transportation and education. The current focus for GenAI
application is on productivity, automation and evolving job roles.

Gartner, Inc. | G00828523 Page 92 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ GenAI is proving its worth in life sciences, manufacturing, finance, law and
entertainment. It is becoming more specialized and optimized for domains such as
coding assistance, scientific discovery, research, diagnostics, legal analysis and
financial modeling. Additionally, 78% of enterprises surveyed by Gartner have
integrated or are planning to integrate the use of GenAI into some areas (see
Technology Spending Drivers, Business Outcomes and Challenges for CIOs Across
Industries for more information).

■ Businesses aim to automate tasks, generate content and enhance customer

experience by integrating GenAI into their processes. Prompt engineering is the main
approach for custom GenAI use cases.

■ Governments, spurred by the GenAI promise, are increasing investments in national

AI strategies.

■ Agentic AI is a top driver of a GenAI value proposition this year due to automation
benefits and combining GenAI with other techniques.

■ Fierce GenAI model competition continues. GenAI providers are introducing model
quality and performance improvements, as well as more sophisticated reasoning
and handling of image and video inputs. Galloping leaderboards list hundreds of
large language models (LLMs), including a variety of smaller models that
demonstrate precision and cost-effectiveness in specific domains and tasks, such as
time series. Distillation, truncation and other methods to derive smaller models from
large ones result in reduced latency and lower costs. Open-source LLMs democratize
access to GenAI and stimulate ecosystem innovation.

■ Technology vendors and service providers compete on GenAI applications and

model offerings, and their enterprise readiness, pricing, infrastructure, safety and
indemnification. Vendors and open-source communities offer better tooling for
training, fine-tuning, evaluation and life cycle.

■ Infrastructure innovations and investments are on the rise. Hyperscalers and some
enterprises are building supercomputing systems that combine innovations in
computational accelerators, high-speed networks and performance-optimized
storage. Meanwhile, innovations like DeepSeek stimulate ideas efficiently with less
advanced chips and lower costs.

Gartner, Inc. | G00828523 Page 93 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles
■ Estimating GenAI's value is challenging, with less than 30% of the AI leaders from
the 2024 Gartner AI Mandates for the Enterprise Survey reporting that their CEOs
praise AI investment returns. Organizations face productivity leakage, where GenAI
adoption doesn't directly yield value.

■ Technical challenges include security, model evaluation, data availability and quality,
and managing compute for inferencing.

■ Low maturity organizations have difficulty in identifying suitable use cases and face
unrealistic expectations for GenAI initiatives.

■ Advanced organizations struggle to find skilled professionals. New users necessitate

GenAI literacy.

■ Governance challenges include hallucinations, bias, fairness and establishing a

governance operating model. Government regulations may impede GenAI initiatives.

■ GenAI licensing and pricing are inconsistent among providers. Pricing remains
confusing and constantly evolving, often catching customers by surprise.

User Recommendations

■ Focus on problems that GenAI can solve effectively. Develop methods to identify
impactful GenAI use cases that align with business objectives and offer tangible
benefits.

■ Design solutions to be loosely coupled with GenAI models to enable flexible model
selection and combinations. Investigate GenAI vendor roadmaps to avoid spending
your own resources on the capabilities that vendors will deliver in the future.

■ Develop an AI-ready data strategy around your GenAI portfolio. Plan to incorporate
your proprietary data into GenAI via retrieval-augmented generation or similar
methods. Ensure data is relevant, clean, and accessible for GenAI models.

■ Invest in AI literacy and talent upskilling for working with GenAI tools and
technologies.

■ Establish GenAI governance operating model, policies, controls and technical

oversight. Consider both your and your vendors’ responsible AI practices.

■ Plan for the cost of running GenAI initiatives, including infrastructure, compute
resources and ongoing maintenance.

Gartner, Inc. | G00828523 Page 94 of 108

This research note is restricted to the personal use of [email protected].

 Sample Vendors

Alibaba Cloud; Amazon Web Services; Anthropic; DeepSeek; Google; Hugging Face; IBM;
Meta; Microsoft; OpenAI

Gartner Recommended Reading

Generative AI: The Basics

Solution Path for Implementing Generative AI Systems

AI Technology Sandwich: A Conceptual Framework for Executing AI

10 Best Practices for Scaling Generative AI Across the Enterprise

Gartner, Inc. | G00828523 Page 95 of 108

This research note is restricted to the personal use of [email protected].

 Climbing the Slope
Cloud AI Services
Analysis By: Jim Scheibmeir, Bern Elliot

Benefit Rating: High

Market Penetration: 20% to 50% of target audience

Maturity: Early mainstream

Definition:

Cloud AI services provide AI model-building tools, APIs for prebuilt services and
associated middleware that enable the building/training, deployment and consumption of
machine learning (ML) and generative AI models running on prebuilt infrastructure as
cloud services. These services include pretrained vision, language and other generative AI
services, and automated ML and fine-tuning to create new models and customize prebuilt
models.

Why This Is Important

The use of cloud AI services continues to increase due to the interest and hype around
generative AI technologies. Vendors have established large language model (LLM) APIs
and solutions with fully integrated MLOps pipelines. The addition of low-code tools aids
ease of use. Applications regularly use AI cloud services in language, vision and tabular
data to automate business processes. Developers are increasingly using both prebuilt and
customized ML models in applications.

Business Impact

Cloud AI services impact business-running applications, allowing developers to enhance

application functions with intelligent automation. Generative AI is the newest category to
these solutions. Data-driven decisioning mandates the inclusion of ML models to add
application functionality. Some AI technologies are maturing, but generative AI includes
less mature capabilities. Cloud AI services enhance applications with models that score,
forecast and generate content, enabling data-driven business operations.

Drivers

■ Demand for conversational interactions: The emergence of generative AI and LLMs

facilitates conversationally enabled applications where users can use LLMs with
data sources to gain insights.

Gartner, Inc. | G00828523 Page 96 of 108

This research note is restricted to the personal use of [email protected].

 ■ Opportunities to capitalize on data investments: The wealth of data from both
internal and third-party sources is highly useful for building predictive ML models
that enable data-driven decision intelligence in applications.

■ Need to meet business key performance indicators (KPIs): There is a need for
businesses to automate processes to improve accuracy, improve responsiveness
and reduce costs by deploying both AI and ML models.

■ Reduced barriers to entry: The ability to use pretrained generative AI models and
fine-tune them has reduced the need for large quantities of data to train models.
Access for developers and citizen data scientists to AI and ML services, due to the
availability of API-callable LLMs, will further expand the use of AI by development
teams.

■ Automated ML as an enabler for custom development: The use of automated ML to

customize packaged services to address specific business needs is much more
accessible and doesn’t require data scientists.

■ Wide range of cloud AI services: A range of specialized providers in the market offer
cloud AI services, including orchestration layers to streamline deployment of
solutions.

■ Emerging AI model marketplaces: New marketplaces should help developers adopt

predictive and foundation models.

Obstacles

■ Developers and citizen data scientists lack understanding about how to adapt cloud
AI services to specific use cases.

■ Grounding generative AI models is a challenge, requiring well-crafted retrieval-

augmented generation (RAG) solutions that often include vector embeddings and
other capabilities to implement. Many cloud AI developer services (CAIDS) providers
are offering these capabilities as part of their generative AI offering.

■ Pricing models for usage-based cloud AI services present a risk for businesses as
the costs associated with their use can accrue rapidly. There is a need for
comprehensive cost modeling tools to address this issue.

■ There is an increased need for packaged solutions that utilize multiple services for
developers and citizen data scientists.

Gartner, Inc. | G00828523 Page 97 of 108

This research note is restricted to the personal use of [email protected].

 ■ There is a lack of skills such as prompt engineering and fine-tuning for developers to
effectively implement these services in a responsible manner.

User Recommendations

■ Improve the chances of success of your AI strategy by experimenting with AI

techniques, including the use of generative AI models such as LLMs and multimodal
models, and other cloud services. Ensure that generative AI models are loosely
coupled as the technology continues to evolve rapidly.

■ Use cloud AI services to build less complex models, giving the benefit of more
productive AI while freeing up your data science assets for higher-priority projects.

■ Empower non-data-scientists with features such as automated algorithm selection,

dataset preparation and feature engineering for project elements. Leverage existing
expertise on operating cloud services to assist technical professional teams.

■ Utilize pretrained generative AI models to allow for rapid prototyping and deployment
of LLM-enabled solutions.

■ Develop cost modeling tools that allow the enterprise to effectively predict both
usage and management costs as AI models are broadly deployed in applications
across the business.

Sample Vendors

Alibaba Cloud; AWS; Baidu; Google; H2O.ai; IBM; Microsoft; NVIDIA; Oracle; Tencent Cloud

Gartner Recommended Reading

Critical Capabilities for Cloud AI Developer Services

Magic Quadrant for Cloud AI Developer Services

Knowledge Graphs
Analysis By: Afraz Jaffri

Benefit Rating: High

Market Penetration: 20% to 50% of target audience

Maturity: Early mainstream

Gartner, Inc. | G00828523 Page 98 of 108

This research note is restricted to the personal use of [email protected].

 Definition:

Knowledge graphs are machine-readable representations of the physical and digital

environments. They include entities (people, companies and digital assets) and their
relationships, which adhere to a graph data model — a network of nodes (vertices) and
links (edges/arcs).

Why This Is Important

Knowledge graphs capture information about the world in a visually intuitive format yet
are still able to represent complex relationships. Knowledge graphs act as the backbone
of a number of products, including search, smart assistants and recommendation
engines. Knowledge graphs support collaboration and sharing, exploration and discovery,
and the extraction of insights through analysis. Generative AI models can be combined
with knowledge graphs to provide context for more accurate outputs in a technique
becoming known as GraphRAG.

Business Impact

Knowledge graphs can drive business impact in a variety of different settings, including:

■ Digital workplace, such as collaboration, sharing and search

■ Automation, such as ingestion of data from content to robotic process automation

■ Machine learning (ML), such as augmenting training data

■ Investigative analysis, such as law enforcement, cybersecurity and risk management

■ Digital commerce, such as product information management and recommendations

■ Data management, such as metadata management, data cataloging and data fabric

Gartner, Inc. | G00828523 Page 99 of 108

This research note is restricted to the personal use of [email protected].

 Drivers

■ The need to complement AI and ML methods that detect only patterns in data (such
as the current generation of foundation models) with the explicit knowledge, rules
and semantics provided by knowledge graphs.

■ The desire to make better use of unstructured data in documents, correspondence,

images and videos, using standardized metadata that can be related and managed
and provide the foundation for AI-ready data.

■ The increased usage of knowledge graphs with large language models to provide
enhanced contextual understanding when answering questions on large quantities
of enterprise data.

■ The increasing awareness of the use of knowledge graphs in consumer products

and services, such as smart devices and voice assistants, chatbots, search engines,
recommendation engines and route planning.

■ The emerging landscape of Web3 applications and the need for data access across
trust networks, leading to the creation of decentralized knowledge graphs to build
immutable and queryable data structures.

■ The need to manage the increasing number of data silos where data is often
duplicated, and where meaning, usage and consumption patterns are not well-
defined.

■ The use of graph algorithms and ML to identify influencers, customer segments,

fraudulent activity and critical bottlenecks in complex networks.

Gartner, Inc. | G00828523 Page 100 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Awareness of knowledge graph use cases is increasing, but business value and
relevance are difficult to capture in the early implementation stages.

■ Moving knowledge graph models from prototype to production requires engineering

and system integration expertise. Methods to maintain knowledge graphs as they
scale — to ensure reliable performance, handle duplication and preserve data quality
— remain immature.

■ Organizations want to enable the ingestion, validation and sharing of ontologies and
data relating to entities such as geography, people and events. However, making
internal data interoperable with external knowledge graphs is a challenge.

■ In-house expertise, especially among subject matter experts, is lacking, and

identifying third-party providers is difficult. Often, expertise resides with vendors of
graph technologies. Skills in scalability and optimization are also hard to acquire.

User Recommendations

■ Create a working group of knowledge graph practitioners and sponsors by

assessing the skills of data and analytics (D&A) leaders, practitioners and business
domain experts. Factors like use case requirements, data characteristics, scalability
expectations, query flexibilities and domain knowledge of knowledge graphs should
be addressed.

■ Run a pilot to identify use cases that need custom-made knowledge graphs. The
pilot should deliver not only tangible value for the business but also learning and
development for D&A staff.

■ Create a minimum viable subset that can capture the information of a business
domain to decrease time to value. Assess the data, both structured and
unstructured, needed to feed a knowledge graph, and follow agile development
principles.

■ Utilize vendor and service provider expertise to validate use cases, educate
stakeholders and provide an initial knowledge graph implementation.

■ Include knowledge graphs within the scope of D&A governance and management.
To avoid perpetuating data silos, investigate and establish ways for multiple
knowledge graphs to interoperate and extend toward a data fabric.

Gartner, Inc. | G00828523 Page 101 of 108

This research note is restricted to the personal use of [email protected].

 Sample Vendors

Altair; Diffbot; eccenca; Fluree; Graphwise; Graphwise (Ontotext); Neo4j; Stardog;

TopQuadrant

Gartner Recommended Reading

How to Build Knowledge Graphs That Enable AI-Driven Enterprise Applications

3 Ways to Enhance AI With Graph Analytics and Machine Learning

Model Distillation
Analysis By: Birgi Tamersoy, Yogesh Bhatt

Benefit Rating: High

Market Penetration: 5% to 20% of target audience

Maturity: Adolescent

Definition:

Model distillation is a technique where a smaller, simpler model (the “student”) is trained
to replicate the behavior of a larger, more complex model (the “teacher”). This is achieved
by having the student model learn from the predictions of the teacher model. The goal is
to retain the performance of the teacher model while benefiting from the efficiency and
reduced resource requirements of the student model.

Why This Is Important

Organizations often face a trade-off between accuracy and efficiency when deploying AI
models. Larger models, or model ensembles, tend to have higher accuracy but need more
computational resources and may have latency issues. Smaller models are more efficient
and require fewer computational resources, but tend to be less accurate when trained
from scratch. Model distillation helps by transferring the information content of large
models into smaller ones, maintaining performance while improving efficiency.

Business Impact

Model distillation can provide:

Gartner, Inc. | G00828523 Page 102 of 108

This research note is restricted to the personal use of [email protected].

 ■ Cost reduction and energy efficiency: Smaller models require fewer computational
resources during deployment, resulting in reduced capital investments, operational
costs and energy consumption.

■ Faster inference times: Efficient models result in quicker inference times, enabling
real-time applications and improving user experience.

■ Compliance and privacy: Smaller models can be deployed on-premises or on a

device, which can help in meeting data privacy regulations.

Drivers

■ Developments in foundation models: As foundation models grow larger and more

capable, they offer the ability to generate high-quality synthetic data, but at the same
time, pose cost challenges for deployment. These developments create both an
enabler and a need for model distillation.

■ Growth in edge computing and mobile applications: The rise of edge computing and
AI-powered mobile applications require AI models that can operate efficiently on
devices with limited computational power. Model distillation helps create smaller
models that are suitable for deployment on such devices.

■ Increased focus on sustainability: As organizations strive to reduce their carbon

footprint, there is a growing emphasis on energy-efficient AI solutions. Model
distillation contributes to sustainability by reducing the computational resources and
energy required for model deployment and operation.

■ Advancements in transfer learning: Developments in transfer learning have

enhanced the ability to transfer knowledge from large models to smaller ones,
improving the effectiveness of model distillation techniques. This allows distilled
models to achieve higher accuracy while remaining efficient.

■ Regulatory pressure and data privacy: Increasing regulations around data privacy
and protection are driving the need for AI models that can be deployed on-premises
or on a device, minimizing data transfer and exposure. Model distillation supports
compliance by enabling the deployment of efficient models in secure environments.

Gartner, Inc. | G00828523 Page 103 of 108

This research note is restricted to the personal use of [email protected].

 Obstacles

■ Loss of model accuracy: Student models may lose some accuracy compared to
teacher models, especially when the student model has significantly fewer
parameters compared to the teacher model.

■ Complexity of the distillation process: The process may require higher levels of
technical expertise, which can be a barrier for some organizations.

■ Limited generalization across domains: Distilled models might not generalize well
across various domains, limiting their applicability in diverse tasks.

■ Dependence on high-quality teacher models: The success of distillation depends on

the quality of the teacher model; any biases, inefficiencies or potential intellectual
property infringement risks can be passed to the student model.

■ Licensing restrictions on teacher models: Some providers impose restrictions on the

use of teacher model outputs, prohibiting their use in the training of other models,
which can limit the feasibility of distillation efforts.

User Recommendations

■ Optimize deployment costs: Use model distillation to reduce computational

resources and lower operational expenses, while maintaining model performance.

■ Prioritize high-quality teacher models: Ensure that your teacher models are well-
aligned for your target task, well-optimized and free from biases. This will improve
the quality of distilled models.

■ Invest in expertise: Invest in skilled personnel or training programs for implementing

effective model distillation. This will ensure that your AI solutions are accurate and
efficient.

■ Align with regulatory compliance: Use model distillation to create efficient models
that can be deployed on-premises or on a device. This approach reduces privacy and
safety concerns and eases regulatory compliance.

Sample Vendors

Amazon Web Services; Google; IBM; Microsoft; OpenAI; Predibase; Snorkel AI

Gartner Recommended Reading

Technical Professionals Need to Track 5 Important LLM Developments

Gartner, Inc. | G00828523 Page 104 of 108

This research note is restricted to the personal use of [email protected].

 Appendixes
See the previous Hype Cycle: Hype Cycle for Artificial Intelligence, 2024

Hype Cycle Phases, Benefit Ratings and Maturity Levels

Table 2: Hype Cycle Phases
(Enlarged table in Appendix)

Gartner, Inc. | G00828523 Page 105 of 108

This research note is restricted to the personal use of [email protected].

 Table 3: Benefit Ratings

Benefit Rating Definition

Transformational Enables new ways of doing business across

industries that will result in major shifts in
industry dynamics

High Enables new ways of performing horizontal

or vertical processes that will result in
significantly increased revenue or cost
savings for an enterprise

Moderate Provides incremental improvements to

established processes that will result in
increased revenue or cost savings for an
enterprise

Low Slightly improves processes (for example,

improved user experience) that will be
difficult to translate into increased revenue
or cost savings.

Source: Gartner (June 2025)

Gartner, Inc. | G00828523 Page 106 of 108

This research note is restricted to the personal use of [email protected].

 Table 4: Maturity Levels
(Enlarged table in Appendix)

Document Revision History

Hype Cycle for Artificial Intelligence, 2024 - 17 June 2024

Hype Cycle for Artificial Intelligence, 2023 - 19 July 2023

Hype Cycle for Artificial Intelligence, 2022 - 8 July 2022

Hype Cycle for Artificial Intelligence, 2021 - 29 July 2021

Hype Cycle for Artificial Intelligence, 2020 - 27 July 2020

Hype Cycle for Artificial Intelligence, 2019 - 25 July 2019

Hype Cycle for Artificial Intelligence, 2018 - 24 July 2018

Hype Cycle for Artificial Intelligence, 2017 - 24 July 2017

Hype Cycle for Smart Machines, 2016 - 21 July 2016

Hype Cycle for Smart Machines, 2015 - 24 July 2015

Hype Cycle for Smart Machines, 2014 - 18 July 2014

Gartner, Inc. | G00828523 Page 107 of 108

This research note is restricted to the personal use of [email protected].

 Recommended by the Authors
Some documents may not be available as part of your current Gartner subscription.

Understanding Gartner’s Hype Cycles

Tool: Create Your Own Hype Cycle With Gartner’s 2024 Hype Cycle Builder
10 Best Practices for Scaling Generative AI Across the Enterprise
When Not to Use Generative AI

Explore Small Language Models for Specific AI Scenarios

How to Interpret and Act on 2025 AI Index Report Insights
Frontier AI Models Like DeepSeek Will Transform I&O Strategies

© 2025 Gartner, Inc. and/or its affiliates. All rights reserved. Gartner is a registered trademark of
Gartner, Inc. and its affiliates. This publication may not be reproduced or distributed in any form
without Gartner's prior written permission. It consists of the opinions of Gartner's research
organization, which should not be construed as statements of fact. While the information contained in
this publication has been obtained from sources believed to be reliable, Gartner disclaims all warranties
as to the accuracy, completeness or adequacy of such information. Although Gartner research may
address legal and financial issues, Gartner does not provide legal or investment advice and its research
should not be construed or used as such. Your access and use of this publication are governed by
Gartner's Usage Policy. Gartner prides itself on its reputation for independence and objectivity. Its
research is produced independently by its research organization without input or influence from any
third party. For further information, see "Guiding Principles on Independence and Objectivity." Gartner
research may not be used as input into or for the training or development of generative artificial
intelligence, machine learning, algorithms, software, or related technologies.

Gartner, Inc. | G00828523 Page 108 of 108

This research note is restricted to the personal use of [email protected].

Table 1: Priority Matrix for Artificial Intelligence, 2025

Benefit Years to Mainstream Adoption

Less Than 2 Years 2 to 5 Years 5 to 10 Years More Than 10 Years

Transformational Composite AI AI Engineering AI-Native Software Artificial General Intelligence

Responsible AI Decision Intelligence Engineering
Embodied AI AI-Ready Data
First-Principles AI World Models
Foundation Models
Generative AI
ModelOps
Multimodal AI

High Edge AI AI Agents AI Simulation

AI Governance Platforms FinOps for AI
AI TRiSM
Causal AI
Cloud AI Services
Knowledge Graphs
Model Distillation
Neurosymbolic AI
Sovereign AI
Synthetic Data

Moderate

Gartner, Inc. | G00828523 Page 1A of 5A

This research note is restricted to the personal use of [email protected].

Benefit Years to Mainstream Adoption

Less Than 2 Years 2 to 5 Years 5 to 10 Years More Than 10 Years

Low Quantum AI

Source: Gartner (June 2025)

Gartner, Inc. | G00828523 Page 2A of 5A

This research note is restricted to the personal use of [email protected].

Table 2: Hype Cycle Phases

Phase Definition

Innovation Trigger A breakthrough, public demonstration, product launch or other event

generates significant media and industry interest.

Peak of Inflated Expectations During this phase of overenthusiasm and unrealistic projections, a flurry of
well-publicized activity by technology leaders results in some successes, but
more failures, as the innovation is pushed to its limits. The only enterprises
making money are conference organizers and content publishers.

Trough of Disillusionment Because the innovation does not live up to its overinflated expectations, it
rapidly becomes unfashionable. Media interest wanes, except for a few
cautionary tales.

Slope of Enlightenment Focused experimentation and solid hard work by an increasingly diverse
range of organizations lead to a true understanding of the innovation’s
applicability, risks and benefits. Commercial off-the-shelf methodologies and
tools ease the development process.

Plateau of Productivity The real-world benefits of the innovation are demonstrated and accepted.
Tools and methodologies are increasingly stable as they enter their second
and third generations. Growing numbers of organizations feel comfortable
with the reduced level of risk; the rapid growth phase of adoption begins.
Approximately 20% of the technology’s target audience has adopted or is
adopting it as it enters this phase.

Years to Mainstream Adoption The time required for the innovation to reach the Plateau of Productivity.

Gartner, Inc. | G00828523 Page 3A of 5A

This research note is restricted to the personal use of [email protected].

Source: Gartner (June 2025)

Table 3: Benefit Ratings

Benefit Rating Definition

Transformational Enables new ways of doing business across industries that will result in
major shifts in industry dynamics

High Enables new ways of performing horizontal or vertical processes that will
result in significantly increased revenue or cost savings for an enterprise

Moderate Provides incremental improvements to established processes that will result

in increased revenue or cost savings for an enterprise

Low Slightly improves processes (for example, improved user experience) that will
be difficult to translate into increased revenue or cost savings.

Source: Gartner (June 2025)

Gartner, Inc. | G00828523 Page 4A of 5A

This research note is restricted to the personal use of [email protected].

Table 4: Maturity Levels

Maturity Levels Status Products/Vendors

Embryonic In labs None

Emerging Commercialization by vendors First generation

Pilots and deployments by industry leaders High price
Much customization

Adolescent Maturing technology capabilities and process Second generation

understanding Less customization
Uptake beyond early adopters

Early mainstream Proven technology Third generation

Vendors, technology and adoption rapidly evolving More out-of-box methodologies

Mature mainstream Robust technology Several dominant vendors

Not much evolution in vendors or technology

Legacy Not appropriate for new developments Maintenance revenue focus

Cost of migration constraints replacement

Obsolete Rarely used Used/resale market only

Source: Gartner (June 2025)

Gartner, Inc. | G00828523 Page 5A of 5A

This research note is restricted to the personal use of [email protected].