The blueprint

to scaling AI
for business
transformation
Transitioning from Pilot
to Production
 Foreword
The journey from pilot to production is fraught
with challenges, but the rewards are substantial.
Organizations that successfully scale their AI
initiatives report significant gains in operational
efficiency, cost savings, and innovation. For
instance, generative AI has enabled companies
like Eneco eMobility, BMW, and ABN AMRO to
achieve remarkable improvements in productivity
and customer service. These success stories
highlight the transformative potential of AI when
integrated into core business functions.
However, scaling AI requires a holistic approach
that encompasses strategic alignment, robust
governance frameworks, a strong data foundation,
and continuous talent development. All these
aspects are intricately linked and interdependent.
Moreover, it is essential to align Gen AI initiatives
with the broader business value chain, establishing
clear KPIs while embedding responsible AI
practices throughout the AI lifecycle to ensure
Mark Oost ethical and secure deployment.
VP, Gen AI Global Lead, Capgemini
[email protected] As we look to the near future, the rise of AI agents
promises to further transform the way businesses
operate by offering several key benefits such as
The strategic importance of AI, particularly sales growth – organizations have seen a 4.4%
generative AI, cannot be overstated. It has moved increase in sales, attributed to the AI agents’ ability
beyond the realm of experimentation and is now to provide personalized recommendations and
integral to enterprise strategies across sectors. optimize sales operations. Building and managing
The findings from Capgemini’s recent research, multi-agent systems is no easy feat, but the key
Harnessing the value of generative AI: 2nd Edition, to success lies in continuously monitoring and
underscore this shift. Our research reveals that 80% evaluating their performance.
of organizations have increased their investment In conclusion, the insights and recommendations
in generative AI over the past year, with nearly presented in this paper provide a valuable guide for
one-quarter integrating this technology into organizations seeking to scale their AI initiatives. It
their operations. This surge in adoption is driven outlines strategies and a comprehensive roadmap
by the tangible benefits generative AI offers, for successful transition to full-scale production.
including a 7.8% improvement in productivity and
a 6.7% enhancement in customer engagement
and satisfaction.

The blueprint to scaling AI for business transformation 2

May 2025

The Blueprint to
Scaling AI for Business
Transformation
Transitioning from Pilot to Production

This document has been

licensed to Capgemini
 Contents
03 Introduction

04 Strategic importance of AI and generative AI

06 Current state of enterprise AI adoption

10 Success factors for AI and generative

AI transition from pilot to production

18 Common pitfalls to avoid when scaling AI

19 ABN AMRO case study

21 Conclusion and future outlook

Copyright © 2025, Everest Global, Inc. All rights reserved. www.everestgrp.com | this document has been licensed to Capgemini
3 The Blueprint to Scaling AI for Business Transformation

Introduction
AI, including generative AI, has moved beyond being an emerging technology confined
to experimentation and is now becoming integral to enterprise strategies. Across
industries, organizations are experiencing measurable gains from pilot projects –
ranging from improved productivity to enhanced customer experiences. The next logical
step is to scale these initiatives from pilot stages to full-scale production to unlock their
transformative potential.

However, transitioning from pilot to production is a complex journey that demands

addressing various challenges and success factors. Enterprises must tackle essential
aspects such as defining an enterprise-wide AI strategy, prioritizing AI use cases,
establishing an effective AI operating model, ensuring data readiness, building AI talent,
implementing a scalable AI technology stack, and integrating responsible AI practices.
Success in scaling AI depends on clearly understanding these factors, avoiding common
pitfalls, and adopting best practices.

This viewpoint delves into the essential aspects of scaling AI, providing insights into:
 Key business drivers behind AI adoption
 The current state of AI and generative AI adoption
 Success factors for scaling AI from pilot to production
 Common pitfalls to avoid while scaling AI initiatives
 Future trends and outlook in AI

This viewpoint provides enterprises with insights into successfully scaling AI from pilot to
production. By exploring key success factors and avoiding potential pitfalls, it aims to
help organizations navigate the complex journey of operationalizing AI and unlocking its
transformative potential. Furthermore, it serves as a guide for businesses looking to
move beyond experimental stages and achieve tangible, enterprise-wide impact with AI.

www.everestgrp.com | this document has been licensed to Capgemini

4 The Blueprint to Scaling AI for Business Transformation

Strategic importance of AI and

generative AI
Economic volatility, evolving customer expectations, regulatory pressures, and rising
competition are causing the business landscape to fundamentally shift. Traditional
transformation approaches have often fallen short, struggling to deliver agility, resilience,
and sustained growth. As a result, organizations are rethinking their strategies – not just
for digital transformation but for holistic business transformation that reshapes
operational models, decision-making, and customer engagement.

Over the past decade, enterprises have invested in modernizing IT systems, cloud
adoption, automation, and data-driven decision-making. While these efforts have laid a
strong foundation, they have often remained siloed within functions, processes, and
systems, limiting their ability to deliver enterprise-wide impact. Now, AI is acting as the
true catalyst, elevating digital transformation into business transformation – one that is
AI-first, data-driven, and outcome-oriented.

This shift requires organizations to break free from fragmented automation and
disjointed AI initiatives. AI, particularly generative AI, presents a unique opportunity to
bridge these gaps, enabling enterprises to move beyond isolated use cases toward
scalable, interconnected, and intelligence-driven transformation. Whether enhancing
customer interactions, optimizing supply chains, or accelerating product innovation,
generative AI is not just enhancing processes but reshaping core business functions.

Enterprises are now transitioning from AI experimentation to AI at scale. The focus is

shifting from pilots to production, with organizations establishing robust governance
frameworks, data ecosystems, and AI-driven decisioning models to ensure enterprise-
wide adoption and measurable impact. The next section explores the key business
imperatives driving this shift.

“It will be unthinkable not to have intelligence

integrated into every product and service.”
– Sam Altman, CEO, OpenAI

www.everestgrp.com | this document has been licensed to Capgemini

5 The Blueprint to Scaling AI for Business Transformation

Key business drivers behind AI adoption

Adopting AI and generative AI is reshaping business priorities across industries.
Organizations increasingly view these technologies as vital to stay competitive, improve
operational efficiency, and drive business outcomes. Below are the key business drivers
steering AI / generative AI adoption:
 Enhancing operational efficiency and reducing turnaround time: AI transforms
operations by automating repetitive tasks and streamlining decision-making. These
technologies enable enterprises to optimize workflows, minimize errors, and
accelerate operations, thereby increasing efficiency and reducing time-to-market for
products and services. For example, AI-driven supply chain optimization through
demand forecasting can significantly improve inventory management accuracy.
Additionally, AI can speed up creative design cycles, prototyping, and reporting
processes, allowing businesses to quickly respond to market demands
 Delivering superior customer experiences: Improving customer experience is a top
priority for businesses, and AI-driven personalization is a game changer. Generative
AI can tailor customer interactions through personalized recommendations, intelligent
chatbots, and content that resonates with target audiences. These capabilities help
businesses foster stronger customer relationships, increase satisfaction, and drive
loyalty
 Increasing employee productivity: AI goes beyond automation, amplifying
employee capabilities through tools such as AI-powered copilots. These copilots draft
content and offer intelligent recommendations, enabling employees to focus on
complex, high-value tasks. For example, coding copilots help developers debug and
code faster, driving productivity across teams
 Maximizing cost savings: AI solutions are instrumental in achieving long-term cost
optimization. From automating product design and optimizing supply chains to
reducing research and development costs, AI technologies help businesses unlock
efficiencies while maintaining high output standards. These cost-saving opportunities
are crucial for reinvesting in innovation and sustainably scaling operations
 Technological maturity: The growing maturity of enabling technologies, such as
robust data and AI infrastructure, scalable cloud solutions, and advanced language
models, has played a key role in AI and generative AI’s widespread adoption. These
technology advances have now made it possible for enterprises to harness the full
potential of these transformative tools, unlocking a wide range of business benefits
 Driving innovation and increasing relevance: AI and generative AI are gaining
prominence as businesses recognize their ability to drive innovation and maintain
competitive relevance. By leveraging these technologies to enhance their products,
services, and processes, organizations can stay ahead, expand market share, and
strengthen their position in an increasingly dynamic, technology-driven landscape

www.everestgrp.com | this document has been licensed to Capgemini

6 The Blueprint to Scaling AI for Business Transformation

Current state of enterprise

AI adoption
As we enter 2025, AI continues to cement its promise as a transformative technology
in the modern business landscape. Based on Everest Group's enterprise buyer survey
conducted in Q2 2024, more than 30% of enterprises plan to increase their spend on
AI services by 10-20% or more in the coming year, while more than 50% have increased
spend by 5-10%, demonstrating a strong commitment to AI investments.

This broader AI momentum has set the stage for generative AI. As we reflect on 2024,
a year dominated by generative AI-related conversations, several questions have
emerged: Is generative AI here to stay, or is it merely a passing trend? Has it moved
beyond curiosity to meaningful enterprise adoption?

One thing is clear – generative AI has been at the center of strategic discussions in
boardrooms globally. According to Everest Group’s Q1 2024 enterprise leaders’
response survey, 83% of organizations worldwide are either actively piloting generative
AI initiatives or have already implemented it in production-grade use cases.

Exhibit 1: Enterprise adoption of generative AI

Source: Everest Group (2025)

61%
actively exploring and
piloting generative AI

22%
have deployed
generative AI for at

15% least one or more

processes
planning to pilot
generative AI soon

2%
have not
discussed it yet

www.everestgrp.com | this document has been licensed to Capgemini

7 The Blueprint to Scaling AI for Business Transformation

Beyond generative AI, enterprises are now exploring the next AI evolution: agentic AI.
Agentic AI systems combine Large Language Model (LLM) capabilities with code, data
sources, and user interfaces to autonomously perform complex tasks and workflows.
This progression signifies a deepening AI integration into enterprise operations, setting
the stage for transformative changes across various industries and business functions.

Generative AI adoption across industries

Enterprises have recognized AI's transformative potential and are increasingly
embedding generative AI into their operations to achieve competitive differentiation and
tangible business value. While 2023 marked generative AI's introduction to businesses,
2024 saw organizations move toward practical implementation. Organizations are
reporting tangible benefits from deploying generative AI. For example, based on public
disclosures, Eneco eMobility enhanced customer service agent productivity, reducing
average wrap-up time by 50%. Similarly, BMW achieved a 30-40% increase in
productivity and improved stakeholder experience.

While generative AI has the potential to impact every industry, different industries are
adopting it at a different pace. In a fast-moving technology landscape such as this,
enterprise consumption patterns and adoption phases are rapidly evolving, influenced by
continuous advances and emerging use cases. Exhibit 2 shows generative AI adoption
patterns across different industries, based on Everest Group’s survey of ~600 enterprise
respondents conducted in Q1 2024.

As per Exhibit 2 on page 8, certain industries such as BFSI and hi-tech and technology
are already extensively using and experimenting with generative AI, compared to other
sectors. A sizable number of enterprises are still experimenting with generative AI to
evaluate these systems’ reliability before amplifying investments.

This industry-specific variation in adopting generative AI can be attributed to several

factors such as current data availability, technical readiness, cost implications,
regulatory and compliance considerations, and generative AI capability requirements.
For example, BFSI and RCPG organizations often have well-structured data repositories
and customer interactions that lend themselves more readily to generative AI-powered
automation and personalization. Conversely, the HLS sector may face more complex
challenges such as data privacy and security concerns and limited training data access
that require a more cautious, phased approach to generative AI adoption.

Irrespective of the industry, the most successful enterprises are those that identify
high-impact, strategic generative AI use cases and establish frameworks to seamlessly
scale these pilots into production-ready solutions.

www.everestgrp.com | this document has been licensed to Capgemini

8 The Blueprint to Scaling AI for Business Transformation

Exhibit 2: Generative AI adoption across industries

Source: Everest Group (2025)
Waiting for Experimenting but Experimenting
generative Not planning Investigating not amplifying Extensively and amplifying
AI maturity to use possibilities investments using investments

BFSI 1% <1% 13% 1% 39% 46%

HLS 2% 2% 26% 6% 24% 40%
RCPG 1% 4% 21% 5% 32% 38%
Automative 4% <1% 15% 13% 20% 48%
Electronics, hi-tech,
<1% 3% 8% 1% 41% 46%
and technology
Media and
<1% <1% 22% 6% 33% 39%
entertainment
Airline, travel, and
3% 3% 25% 10% 13% 48%
transport
Energy and utilities 2% 2% 19% 2% 17% 58%
Others <1% 5% 20% 5% 16% 53%

Key business transformation areas where AI is transitioning from

pilot to production
Enterprises are leveraging AI and generative AI across a wide spectrum of transformation
areas to drive efficiency and enhance stakeholder experiences.

In the front office, AI is transforming customer experience and sales and marketing.
By automating tasks such as summarizing complaints, retrieving data, and crafting
personalized responses to queries, it improves agent productivity and customer
experience. Generative AI optimizes sales campaigns and customer targeting by creating
tailored content and experiences for individual personas, making front office operations
more efficient, streamlined, and customer-focused.

In the middle office, AI is revolutionizing risk management, performance analysis, and data
management. Risk management benefits from automated regulatory compliance reporting,
fraud detection, and tax reporting, reducing manual effort and improving accuracy. In
performance analysis, AI enhances decision-making through predictive and prescriptive
insights, conversational analytics, and dynamic reporting, enabling teams to uncover trends
and make informed decisions. In data management, generative AI accelerates data-related
operations such as schema generation, data migration, and synthetic data generation.

In the back office, AI is driving transformation across finance and accounting, HR, IT
operations and software development, and supply chain management:

 In finance and accounting, AI enhances intelligent document processing by automating

tasks such as generating and summarizing financial reports and extracting key details
from financial documents, significantly improving accuracy and efficiency
www.everestgrp.com | this document has been licensed to Capgemini
9 The Blueprint to Scaling AI for Business Transformation

 In HR, generative AI streamlines recruitment, Learning and Development (L&D), and

talent management by automating screening resumes, generating job descriptions,
creating personalized training content, and providing tailored responses to employee
queries
 In supply chain management, AI enhances warehouse operations and inventory
management through near real-time analytics, enabling improved decision-making,
optimized logistics, and increased operational efficiency
 In IT operations and software development, AI automates detecting and resolving
incidents, monitoring systems, assisting code, and generating test cases, minimizing
downtime and improving efficiency

Exhibit 3 highlights key transformation areas in various functional domains within

enterprises.

Exhibit 3: Key transformation areas across functional domains

Source: Everest Group (2025) [NOT EXHAUSTIVE]

Customer Customer support Agent assistance Knowledge

experience management

Sales and Campaign Content creation and Sales enablement

marketing management personalization and customer insights
and analytics

Risk Regulatory compliance Fraud detection Knowledge

management and reporting management

Finance and Financial reporting Financial document Budgeting and cash

accounting processing flow forecasting

HR Employee Personalized L&D Talent acquisition

engagement and and recruitment
communication

IT operations Application IT help desk and Incident and data

and software development support management
development

Supply chain Procurement Supply chain planning Transportation

management management

Across these transformation areas, factors such as the varying degrees of complexity,
data availability, abundance of potential applications, and need for automation are
shaping the pace and scale of AI adoption. For example, customer experience and sales
and marketing are witnessing the highest AI adoption due to their wide range of use
cases and relatively lower complexity.

Exhibit 4 illustrates some of the enterprises that have successfully transitioned AI

solutions from pilot to production across different transformation areas.

www.everestgrp.com | this document has been licensed to Capgemini

10 The Blueprint to Scaling AI for Business Transformation

Exhibit 4: Success stories across transformation areas

Source: Everest Group (2025)

Customer support Financial document

and content creation processing
Deployed generative AI solutions for Launched DocLLM, an AI solution to
shopping assistance and AI-driven process complex enterprise documents
product descriptions (e.g., forms, invoices)
Carrefour J. P. Morgan

Application development Knowledge management

Implemented a generative AI coding Implemented a generative AI assistant

assistant to support developers in to support financial advisors and staff
coding tasks in performing research tasks
Goldman Sachs Morgan Stanley

Success factors for AI and

generative AI transition from pilot
to production
Enterprises across certain domains and industries are successfully transitioning AI use
cases from pilot to production. Successful organizations have scaled these initiatives by
addressing critical factors. Exhibit 5 outlines some of the key success factors to scale AI
from pilot to production.

“The most important thing we need to do is

ensure that AI aligns with the business’
broader goals and is solving real problems
for customers.”
– Satya Nadella, CEO, Microsoft

www.everestgrp.com | this document has been licensed to Capgemini

11 The Blueprint to Scaling AI for Business Transformation

Exhibit 5: Success factors to scale AI from pilot to production

Source: Everest Group (2025) [NOT EXHAUSTIVE]

Success factor Enterprise-wide AI strategy Operating model for AI Responsible AI

Activities Align AI strategy with business Select the right operating Establish a robust AI
objectives model governance framework

Identify and prioritize AI use cases Embrace agile delivery Mitigate risks proactively
models across AI life cycle
Define measurable goals and KPIs
Empower employees with
Build strong AI leadership
AI risk awareness
Define budgeting and funding
processes

Success factor AI talent Data foundation AI tech stack

Activities Assess the roles and skills Ensure high-quality Select a suitable generative
required for AI annotated data availability AI sourcing strategy

Provide tailored L&D programs Improve data quality Decide on open-source vs.
practices proprietary models
Acquire and retain top AI talent
Implement a modern data Contextualize AI models
platform
Enable reusability across
AI solutions

Enterprise-wide AI strategy
 Align AI strategy with business objectives: An AI strategy must integrate with the
organization’s overall business strategy and objectives to ensure AI investments and
initiatives deliver value. Engage senior leadership to secure buy-in and a top-down
mandate. Leaders must understand how AI creates business value and remain
committed to its integration and success

 Identify and prioritize use cases: Strategically identify and prioritize use cases once
a transformation area has been selected (as highlighted in Exhibit 3). Assess potential
use cases based on business value, feasibility, and risk to ensure effective
prioritization. Eliminate non-performing pilots and scale high-impact initiatives,
minimizing risks and driving sustainable AI transformation. The x-axis in Exhibit 6
represents the ease of adopting generative AI, influenced by factors such as
regulation, data sensitivity, task criticality, and associated risk, while the y-axis
indicates how valuable generative AI is for a task, based on reasoning complexity,
need for generative capabilities, among others. Enterprises should start with use
cases in the Accelerate quadrant and then continue with others as they mature

www.everestgrp.com | this document has been licensed to Capgemini

12 The Blueprint to Scaling AI for Business Transformation

Exhibit 6: Framework to identify and prioritize use cases

Source: Everest Group (2025) [NOT EXHAUSTIVE]

High
Need for generative capabilities
Creating text, video, images, audio

Measures amenability of area to

Reasoning

generative AI application
Potential and impact
Nature of reasoning involved – Evaluate Accelerate
more complex reasoning makes
tasks less amenable to fully adopt
generative AI

Dynamism of knowledge base

Measure of the rate of change of
the knowledge based on which a
task is performed
Wait Educate
Data availability
Data access to train generative AI Low
(foundational model fine-tuning or
net new models) Low Ease of adoption High
Measures roadblocks or lack thereof in
the path to adoption

Regulation Data sensitivity Criticality Risk factor

Degree of regulation – Nature of data to be processed The cost/penalty of task Degree of risk associated
highly regulated businesses require – highly sensitive / private data failure – highly critical tasks with the use case. Higher
clarity on IP laws and explainability, may not be appropriate for and processes require the risk, lower the ease of
making them less likely to adopt generative AI use cases higher human touch adoption, and vice versa
generative AI at scale

 Define measurable goals and KPIs: Establish clear KPIs aligned with business
objectives to track RoI of AI investments. For instance, in customer service, track
metrics such as average handling time and customer satisfaction scores, while in
sales and marketing, monitor campaign RoI and conversion rates. Continuously
monitor the set KPIs and communicate progress to sustain leadership support and
alignment
 Build strong AI leadership: Appoint an expert with cross-functional influence to
oversee the organization's AI initiatives. Establish a steering committee to provide
strategic direction, coordinate AI efforts, and ensure alignment across the enterprise.
Together, these leaders and teams are accountable for achieving AI-driven business
objectives
 Define budgeting and funding: Establish clear budgeting and funding processes to
sustain AI initiatives. Proposals for AI investments must demonstrate measurable
business value to secure board approval. Ensure funding allocations remain agile to
adapt to changing business needs and evolving AI demands

www.everestgrp.com | this document has been licensed to Capgemini

13 The Blueprint to Scaling AI for Business Transformation

Operating model for AI

 Select the right operating model: The operating model defines how people,
processes, and technology are deployed to achieve AI / generative AI goals, directly
impacting AI initiatives’ scalability and effectiveness. Enterprises typically adopt one of
three approaches – centralized, decentralized, or hybrid (federated). Each model
offers unique advantages and challenges, making it essential to align the selection
with organization’s current AI maturity, risk tolerance, governance requirements, and
the need for speed and customization

Exhibit 7: Various AI / generative AI operating models

Source: Everest Group (2025) [NOT EXHAUSTIVE]

Centralized Decentralized Federated

Description

Generative AI capabilities Generative AI fully owned Combines centralized

centralized under CoE and managed by individual oversight with BU-driven
BUs generative AI development
When to adopt

Early stages of adoption When BUs have strong When BUs have some
and in highly regulated generative AI expertise and generative AI expertise
industries require customization to but still need central team
address BU-specific needs support
Benefits

Consistent enterprise-wide Faster customized Balances central control

deployment, robust risk generative AI development, with individual BUs' agility
management fosters innovation within
individual BUs
Challenges

Limited adaptability Inconsistent governance and Coordination complexities

to specific BU needs, security and compliance, risk between the central team and
scalability challenges as of duplicating efforts across BUs, potential inefficiencies
demand grows BUs if responsibilities are not
well-defined

 Embrace agile delivery models: Generative AI’s dynamic nature calls for agile
delivery methods that enable iterative development, rapid experimentation, and
continuous improvement. Key aspects include adopting Continuous Integration and
Deployment (CI/CD) practices, incorporating feedback loops, and monitoring
performance to make timely adjustments

www.everestgrp.com | this document has been licensed to Capgemini

14 The Blueprint to Scaling AI for Business Transformation

Responsible AI
 Establish a robust AI governance framework: To responsibly deploy and scale
generative AI, establish a comprehensive AI governance and risk management
framework. Form an AI ethics council, a cross-functional team responsible to identify
AI risks, shape policies, and drive risk mitigation strategies. For example, Sony Group
established the AI Ethics Committee to provide expertise across all its business units.
On top of that, clearly define roles and ownership across the AI life cycle to ensure
accountability and informed decision-making. The framework should remain adaptable
to evolving regulations, such as the EU AI Act, to ensure compliance while
accommodating new guidelines
 Mitigate risks proactively across AI life cycle: Integrating risk mitigation measures
across the entire generative AI life cycle is essential for responsible and sustainable
deployment. During the development stage, teams need to establish clear processes
to address key risks such as data privacy, security, algorithmic biases, trust and
performance, and ethical concerns. Techniques such as Reinforcement Learning from
Human Feedback (RLHF) and data grounding can enhance trust and performance by
improving model reliability and accuracy. Differential privacy and federated learning
approaches can mitigate privacy and security risks, while bias auditing and data
fairness tools help address algorithmic biases and ensure ethical outcomes. Red
teaming plays a key role in identifying vulnerabilities and providing a robust defense
against potential threats
 Empower employees with AI risk awareness: Fostering an AI risk awareness
culture is essential for responsible generative AI development. Employees, from
engineers to business users, must receive training on potential risks and mitigation
strategies. This empowers them to identify and escalate concerns, ensuring proactive
risk management throughout the AI life cycle

“We need to have a better system of checks

and balances to test AI for bias and fairness.”
– Timnit Gebru, Founder, The Distributed AI Research Institute

AI talent
 Assess the roles and skills AI requires: As organizations strive to scale AI,
they must define the roles and specialized skills required across the AI life cycle.
Successful AI implementation requires investing in specific talent personas and skill
sets. The skills mapping in the exhibit below showcases some of the roles / talent
personas and skills growing in demand due to generative AI. However, these roles
do not need to form entirely new categories; rather, they represent a skill and
expertise evolution within the current workforce, enabling smooth and secure
generative AI adoption

www.everestgrp.com | this document has been licensed to Capgemini

15 The Blueprint to Scaling AI for Business Transformation

Exhibit 8: Top roles and skills for AI / generative AI development

Source: Everest Group (2025) [NOT EXHAUSTIVE]

Role Skills
Generative Generative AI developers LLM development, LLM fine-tuning, API integration
AI-specific or AI/ML engineers
LLMOps specialists Model monitoring, benchmarking, automating pipelines, CI/CD for AI
Prompt engineers Prompt engineering, prompt debugging
Model validators/annotators Data annotation, labeling, response scoring, response validation

D&A Responsible AI and AI audit and risk assessments, compliance knowledge

and AI explainability specialists
Data scientists Statistical modeling, machine learning, data wrangling,
feature engineering
Data and AI architects Data modeling, data governance, cloud platforms

Data engineers ETL, data warehousing, big data, SQL

Consultants Industry and domain experts, Industry expertise, data and AI strategy development
AI consultants

 Provide tailored L&D programs: Building quality AI talent requires investing in

tailored L&D initiatives. This includes creating AI literacy programs and role-specific
trainings to ensure targeted upskilling. Hands-on activities, such as hackathons, can
empower employees to experiment with AI and build competencies in AI / generative
AI applications
 Acquire and retain top AI talent: To address skill gaps, enterprises can combine
upskilling current employees with recruiting new talent. Targeted hiring through
campus recruitment, lateral hiring, or strategic acquisition brings in external expertise.
Retention is equally important – by fostering a culture of innovation, providing
continuous learning opportunities, and recognizing employee contributions,
organizations can effectively nurture and retain top generative AI talent

Data foundation
 Ensure high-quality annotated data availability: Implementing robust data
annotation processes is vital to create high-quality, unbiased training data for AI
models. This involves implementing robust QA/QC processes such as multi-level
reviews and advanced algorithms to detect and mitigate biases. Augmenting datasets
with synthetic data can help create more diverse training data, improving model
accuracy and fairness
 Improve data quality practices: Implement robust data quality management
processes to assess data accuracy, completeness, and consistency. This includes
data profiling, validation, and cleansing techniques. Establish clear data quality
standards and SLAs to proactively address data quality issues. Automate data
quality checks and monitoring, using tools that can detect anomalies, outliers, and

www.everestgrp.com | this document has been licensed to Capgemini

16 The Blueprint to Scaling AI for Business Transformation

other quality issues in real time. Invest in data observability capabilities to gain
end-to-end visibility into data pipelines’ health

 Implement a modern data platform: Deploy a modern, cloud-based data platform to

support scalable AI/ML workloads. Adopt a modular, cloud-native data architecture to
integrate diverse data sources and effectively process structured and unstructured
data. To better support generative AI use cases, explore technologies such as vector
databases to manage embeddings. Leverage cloud-based data management
services, such as data lakes, warehouses, and lakehouses, to establish a unified,
well-governed data foundation. Empower data and AI teams with self-service access
through intuitive, user-friendly data discovery and cataloging capabilities

“If you do not have a strong data foundation, you are

really going to struggle to be able to do anything beyond
well clever parlor tricks with generative AI.”
– Tom Godden, Enterprise Strategist, CXO Advisor, AWS

AI tech stack
 Select a suitable AI sourcing strategy (build versus buy versus partner
approach): Enterprises are broadly exploring the build, buy, and partner approaches
to develop their AI capabilities. Exhibit 9 outlines these approaches.

Exhibit 9: Overview of AI sourcing strategies

[NOT EXHAUSTIVE]
Source: Everest Group (2025)

AI development
approach Build Buy Partner
Developing AI capabilities Involves adopting Engaging external
entirely in-house, including generative AI through tech/service providers
models, using internal existing enterprise to develop generative
resources applications or native AI capabilities
solutions

Ideal scenario Enterprises aiming to Enterprises seeking Enterprises with limited

to use develop customized AI quick benefits with low expertise requiring tailored
products data and AI maturity generative AI solutions

Challenges High costs, requires expert Limited competitive Data privacy and security
in-house AI talent differentiation and concerns, integration complexities,
customization external dependencies

www.everestgrp.com | this document has been licensed to Capgemini

17 The Blueprint to Scaling AI for Business Transformation

In the partner approach, selecting the right external provider is a vital step. Enterprises
must follow structured processes to source, evaluate, and contract partners to align with
their needs. By understanding each approach’s strengths and limitations, enterprises
can choose the most suitable AI implementation strategy that balances speed, control,
and customization for successful outcomes.

 Decide on open-source versus proprietary (closed-source) models: Enterprises

must carefully evaluate the pros and cons of open-source and proprietary models
when building their AI technology stack. Open-source models, such as GPT-J and
LLaMA, are highly customizable and cost-efficient. However, they require significant
in-house expertise for tasks such as pre-processing data and tuning models.
Proprietary models from providers such as OpenAI and Google offer better
performance and reduce Time-To-Value (TTV), but come at higher costs and potential
vendor lock-in risks. Enterprises must consider factors such as model accuracy, data
privacy, compliance, and total cost of ownership when choosing between open-source
and proprietary models. The ideal choice depends on an enterprise's goals,
resources, and technical capabilities. Open-source models may suit organizations
with strong data science teams and customization needs, while proprietary models
may be better suited for enterprises seeking faster TTV and reliable performance
 Contextualize AI models: Enterprises can extract maximum value from AI models
by contextualizing them to their specific requirements. Techniques such as retrieval-
augmented generation, prompt engineering, RLHF, and model fine-tuning significantly
help in this context and enhance model relevance and accuracy. Integrating LLMOps
frameworks ensures continuous model performance monitoring and validation,
enabling ongoing adjustments to align with evolving business objectives and adapt to
new data. This approach helps create AI solutions that are not only accurate but also
agile and responsive to business changes
 Enable reusability across AI solutions: Enterprises can accelerate generative AI
implementation by focusing on developing reusable and modular capabilities. By
analyzing high-priority use cases and identifying common functions, organizations can
build interoperable components that serve multiple needs. To further enhance
reusability, enterprises can also leverage pre-built templates and frameworks
available from AI providers or specialized marketplaces. This modular approach
allows enterprises to mix and match pre-built assets, reducing the need to reinvent
core functionalities for each new AI initiative. Embedding this reusability into the AI
foundation helps create a flexible AI technology stack that supports faster innovation
and improved RoI

“High-performing organizations are nearly three times as

likely to build an AI foundation that enables reuse across
solutions, driving speed and scalability.”
– Walter L, Director, Strategic Planning and Innovation, Microsoft

www.everestgrp.com | this document has been licensed to Capgemini

18 The Blueprint to Scaling AI for Business Transformation

Common pitfalls to avoid when

scaling AI
While success factors lay the groundwork for a strong AI implementation, organizations
need to be mindful of potential challenges that can hinder progress:
 Risk of drifting away from strategic business goals: Organizations risk
misaligning AI projects with strategic objectives by prioritizing low-hanging fruits or
quick wins over broader business transformation. While quick wins deliver immediate
value, an over reliance on them can lead to fragmented AI efforts that fail to address
long-term business objectives. Companies must establish a governance framework
linking AI initiatives to measurable outcomes and balance short-term gains with long-
term strategic priorities. By treating AI as a strategic enabler, organizations can drive
meaningful business transformation and generate tangible business impact
 Uncontrolled costs: AI projects often see costs spiraling out of control due to
extensive data usage and model interactions. The actual cost structure is
multilayered, with model development representing only a small fraction of the total
investment, while change management, operational costs, and continuous
optimization consume significantly more resources. Organizations often overlook
these ongoing expenses, focusing primarily on initial development costs. Effectively
managing these costs is essential to scale AI programs. A holistic financial framework
is essential, along with KPIs to systematically track and manage cost drivers
 Challenges pertaining to AI integration: Integrating AI into existing IT
infrastructures is challenging, particularly when dealing with legacy systems.
Organizations often underestimate the effort required to align AI technologies with the
existing systems, leading to inefficiencies and delays. To mitigate integration risks,
enterprises can adopt a phased approach, starting with smaller projects for real-world
testing to identify and address these issues early. The key is to view AI integration as
an iterative, adaptive process rather than a one-time technical implementation
 Neglecting continuous monitoring and evaluation: Enterprises may underestimate
AI’s evolving nature, treating it as a one-time solution instead of an ongoing process.
With rapid technology advances and changing business needs, AI systems must be
continuously refined and aligned with organizational goals. Organizations must foster
a culture of continuous learning and improvement, establish regular performance
reviews, and create flexible feedback loops for adjustments and system optimizations
 Inability to balance innovation with ethical considerations: As organizations push
AI innovation, they often struggle to balance cutting-edge capabilities with ethical
considerations. The drive to use LLMs can overshadow concerns such as fairness,
transparency, accountability, and unexpected outputs. To mitigate these risks,
organizations must integrate ethical frameworks into the AI development process from
the outset. This integration includes implementing safeguards such as toxicity and
bias detection, sensitive data masking, transparency in AI decision-making, and
regular ethical audits. By balancing innovation with strong oversight, enterprises can
ensure their AI systems remain ethical and trustworthy

www.everestgrp.com | this document has been licensed to Capgemini

19 The Blueprint to Scaling AI for Business Transformation

ABN AMRO case study

Context: ABN AMRO offers services in conversations to prevent misinformation
commercial, retail, and private banking, or policy misalignment required
serving over 350,000 commercial clients additional safeguards.
and 5 million retail clients. Recognizing
Approach: ABN AMRO initially
the importance of seamless customer
deployed its chatbot in 2018. In 2023,
experiences, ABN AMRO’s customer
with a focus on enhancing client
interactions department launched a
interactions and leveraging generative
chatbot as part of its first AI-driven
AI, the bank reevaluated its technology
transformation. Initially leveraging a rules-
landscape. As a result, it transitioned to
based engine and basic NLP models, the
a combination of Microsoft Copilot and
bank automated customer interactions but
Azure services. This shift was driven as
faced limitations as responses were
ABN AMRO identified generative AI as
scripted and did not meet evolving
an opportunity to boost customer
customer expectations adequately.
experience and make digital interactions
Over time, as the need for more context feel more like one-on-one personal
aware and personalized solutions grew, conversations. To facilitate seamless
the bank wanted to build on this migration and implementation, ABN
foundation by incorporating generative AI AMRO partnered with Capgemini,
into its strategy to enhance both customer leveraging its expertise in AI deployment
support and internal operations. and transformation.

Challenge: Despite management and To ensure successful deployment,

stakeholders recognizing generative AI as business and IT teams aligned to secure
a key industry trend and impact area, ABN funding and establish a structured
AMRO had to prioritize strict adherence to implementation roadmap. Adopting a
internal policies, regulatory requirements, phased approach, the bank introduced a
and compliance standards in its first generative AI-powered chatbot and a
generative AI deployment toward clients. summarization tool to enhance AI-driven
Given the complexities of full-scale client assistance.
production, the bank undertook a multi-
Given generative AI’s inherent
month evaluation to reinforce confidence
challenges, ABN AMRO took a cautious
among key decision-makers.
approach, moving beyond traditional AI
Unlike traditional chatbots that relied on risk management measures to focus on
intent recognition and scripted dialogues, conversation control and AI-generated
the major challenge with generative AI response validation. This reinforced the
was ensuring that AI-generated responses need for robust AI governance
were aligned with the bank’s policies, frameworks to continuously monitor
factually accurate, and communicated in chatbot outputs, advanced validation
the right tone. While regulatory mechanisms to fact-check AI responses
compliance, AI governance, and bias against internal knowledge bases, and
mitigation were standard considerations, stress-testing initiatives such as
the need to control AI-driven

www.everestgrp.com | this document has been licensed to Capgemini

20 The Blueprint to Scaling AI for Business Transformation

breakathons to evaluate AI-generated The bank anticipates high client

conversations under various scenarios satisfaction post-production, driven by
and identify potential failures before personalized and human-like
production deployment. conversations facilitated by the
enhanced chatbot capabilities.
ABN AMRO partnered with Capgemini to
Additionally, by automating routine
define the generative AI strategy for the
tasks, human advisors can now focus on
customer interactions department.
higher-value activities such as tailored
Together, they established key
client advisory and strategic decision-
governance elements, including AI
making, leading to improved service
guardrails, to set explicit boundaries for
quality and operational efficiency.
responses and iterative feedback loops to
enhance AI accuracy and alignment. Future outlook: ABN AMRO aims to
Additionally, ABN AMRO continues to further expand its AI initiatives by
leverage Capgemini’s AI talent pool to integrating generative AI capabilities to
scale AI initiatives and build internal automate responses and execute tasks,
capabilities through targeted L&D enhancing customer personalization and
sessions. engagement.

Business outcomes: Prior to generative The bank envisions a future where every
AI implementation, ABN AMRO’s chatbots customer has access to a digital
automated over 2 million conversations personal banker, enabling real-time
annually, ensuring that no more than 30% insights, proactive financial guidance,
required human intervention. With and seamless banking experiences. At
generative AI integration, the impact is the same time, every employee or
expected to be even greater: human agent benefits from an AI-
 Improved first-time resolution: The powered knowledge search and other
percentage of conversations requiring AI-driven assistance.
human support is projected to decrease To achieve these goals and scale at an
from 30% to 23% accelerated pace, ABN AMRO looks
 Enhanced summarization: The new forward to strengthening its partnership
AI-powered summarization tool is with Capgemini, leveraging its expertise
expected to generate summaries for in AI transformation and advanced
800-900 advisors, producing over automation technologies.
40,000 summaries each month,
improving efficiency, and expanding
access to essential information

“Having the right technology services partner and

leveraging its expertise is essential to successfully
transition a PoC into full-scale production.”
– Bobby van Groningen, IT Engineering Lead, ABN AMRO

www.everestgrp.com | this document has been licensed to Capgemini

21 The Blueprint to Scaling AI for Business Transformation

Conclusion and future outlook

As enterprises navigate the complex journey of scaling AI from pilot to production,
success demands a comprehensive and strategic approach that transcends
implementation. Organizations must weave together multiple key elements: strategic
alignment that ties AI initiatives directly to business objectives, robust governance
frameworks that ensure deploying responsible and ethical AI, continuous talent
development that empowers employees with emerging skills, and innovation that
remains adaptive to rapid market changes. Organizations that effectively address these
factors while avoiding common pitfalls are well-positioned to unlock significant value and
gain a competitive edge.

AI’s future is rapidly evolving, with new advances reshaping how businesses operate.
Smaller, more efficient language models are making AI more accessible and affordable,
with companies and start-ups proving that powerful AI does not need huge
infrastructure. Additionally, new AI models are being introduced and existing ones are
being enhanced to provider better reasoning capabilities.

One of the most notable shifts is the rise of agentic AI, which goes beyond traditional
automation by enabling systems to make independent decisions, adapt in real time, and
execute complex workflows with minimal human intervention. Unlike conventional rules-
based automation, agentic AI leverages multi-agent collaboration, self-learning
mechanisms, and generative AI models to drive end-to-end process automation. This
shift is set to transform industries and functions – revolutionizing risk assessment,
compliance, customer service, and enterprise operations. As organizations move away
from fragmented tools toward intelligent automation platforms, agentic AI emerges as
the key to unlocking autonomous, goal-driven execution. Its ability to self-optimize and
handle dynamic business environments makes it not just an evolution but the next
frontier of AI-driven transformation.

With these advances, AI adoption is set to rapidly grow across different industries and
business functions. It will transform the nature of work by automating repetitive tasks,
enhancing workflows, and enabling better decision-making. Business functions will see
streamlined operations, reduced costs, and a shift toward higher-value activities,
redefining how enterprises approach work.

www.everestgrp.com | this document has been licensed to Capgemini

 Everest Group is a leading research firm helping business leaders make
confident decisions. We guide clients through today’s market challenges
and strengthen their strategies by applying contextualized problem-
solving to their unique situations. This drives maximized operational
and financial performance and transformative experiences. Our deep
expertise and tenacious research focused on technology, business
processes, and engineering through the lenses of talent, sustainability,
and sourcing delivers precise and action-oriented guidance. Find further
details and in-depth content at www.everestgrp.com.

This study was funded, in part, by Notice and Disclaimers

Capgemini Important information. Please read this of an offer to purchase any securities or
notice carefully and in its entirety. By instruments from any entity. Nothing from
accessing Everest Group materials, Everest Group may be used or relied upon in
For more information about evaluating the merits of any investment. Do not
products or services, you agree to
Everest Group, please contact: base any investment decisions, in whole or
Everest Group’s Terms of Use.
part, on anything provided by Everest Group.
+1-214-451-3000 Everest Group’s Terms of Use, available at
Everest Group materials, products and/or
www.everestgrp.com/terms-of-use, is hereby
[email protected] services represent research opinions or
incorporated by reference as if fully reproduced
viewpoints, not representations or statements
herein. Parts of the Terms of Use are shown
of fact. Accessing, using, or receiving a grant
below for convenience only. Please refer to the
of access to Everest Group materials, products
link above for the full and official version of the
and/or services does not constitute any
Terms of Use.
recommendation by Everest Group to (1) take
For more information about Everest Group is not registered as an any action or refrain from taking any action or
this topic please contact the author(s): investment adviser or research analyst (2) enter into a particular transaction. Nothing
with the U.S. Securities and Exchange from Everest Group will be relied upon or
Anish Nath, Practice Director Commission, the Financial Industry Regulation interpreted as a promise or representation as
[email protected] Authority (FINRA), or any state or foreign to past, present, or future performance of a
(non-U.S.) securities regulatory authority. For business or a market. The information
Abhiram Srivatsa, Senior Analyst the avoidance of doubt, Everest Group is not contained in any Everest Group material,
providing any advice concerning securities as
[email protected] product and/or service is as of the date
defined by the law or any regulatory entity or prepared and Everest Group has no duty or
an analysis of equity securities as defined by obligation to update or revise the information
the law or any regulatory entity. All properties, or documentation.
assets, materials, products and/or services
Everest Group collects data and information
(including in relation to gen AI) of
from sources it, in its sole discretion, considers
Everest Group are provided or made
reliable. Everest Group may have obtained
available for access on the basis such is for
data or information that appears in its
informational purposes only and provided
materials, products and/or services from the
“AS IS” without any warranty of any kind,
parties mentioned therein, public sources, or
whether express, implied, or otherwise,
third-party sources, including data and
including warranties of completeness,
information related to financials, estimates,
accuracy, reliability, noninfringement,
and/or forecasts. Everest Group is not a
adequacy, merchantability or fitness for a
certified public accounting firm or an accredited
particular purpose. All implied warranties
auditor and has not audited financials.
are disclaimed to the extent permitted by
Everest Group assumes no responsibility
law. You understand and expressly agree
for independently verifying such information.
that you assume the entire risk as to your
use and any reliance upon such. Companies mentioned in Everest Group
materials, products and/or services may be
Everest Group is not a legal, tax, financial, or
customers of Everest Group or have interacted
investment adviser, and nothing provided by
with Everest Group in some other way,
Everest Group is legal, tax, financial, or
including, without limitation, participating in
investment advice. Nothing Everest Group
Everest Group research activities.
provides is an offer to sell or a solicitation

www.everestgrp.com
About Capgemini
Capgemini is a global business and technology transformation partner, helping organizations to accelerate
their dual transition to a digital and sustainable world, while creating tangible impact for enterprises and
society. It is a responsible and diverse group of 340,000 team members in more than 50 countries. With
its strong over 55-year heritage, Capgemini is trusted by its clients to unlock the value of technology to
address the entire breadth of their business needs. It delivers end-to-end services and solutions leveraging
strengths from strategy and design to engineering, all fueled by its market leading capabilities in AI,
generative AI, cloud and data, combined with its deep industry expertise and partner ecosystem. The
Group reported 2024 global revenues of €22.1 billion.

www.capgemini.com

Copyright © 2025 Capgemini. All rights reserved.