RIBA AI Report

RIBA AI Report 2024

2024

1
RIBA AI Report 2024

Contents
Foreword 03
Muyiwa Oki, RIBA President 2023 - 2025
RIBA AI, generative design, and data
RIBA will continue to monitor developments
Digital twin technologies: a major opportunity 04 and provide expert opinion and guidance
Philip D. Allsopp. D.Arch., M.S.(Public Health), RIBA, about emerging and developing technology,
CSBA, CEO, ORBIS Dynamics, Inc and the profession.

An expert advisory group (EAG) has been

The current state, promise, and near-term future for AI in architecture 07 set up to help steer this work.
Amy Bunszel, Executive Vice President, AEC Solutions at Autodesk
The RIBA AI, generative design, and data EAG
Architects and artificial intelligence 10 is Co-Chaired by Phil Allsopp and Nenpin
Dale Sinclair, Head of Digital Innovation at WSP Dimka, with a membership of Greta Jonsson,
Maryam al Irhayim, and Des Fagan. RIBA
staff support is provided by Alex Tait and
RIBA AI survey: findings 12 Adrian Malleson.
Adrian Malleson, Head of Economic Research
Royal Institute of British Architects This group is building on the findings of the
RIBA AI survey, presented here, to look at the
Adoption of AI and its use within the architecture industry 27 broader, ethical, professional practice, and
Jaina Valji, Architect & Founder at Copy and Space competitive implications of the widespread
integration of advancing digital technologies
within architects’ practice.
The computable and the incomputable 30
Tom Holberton, AI Researcher and Lecturer, UCL This EAG brings a wide range of expertise and
experience in the fields of generative design
systems, architectural practice, education and
research, and public policy to examine both the
opportunities and threats to the profession of
architecture. The group will be using case
studies to demonstrate the potential of these
rapidly emerging technologies for equipping
architects to play far more prominent and
sought-after roles. These range from shaping
climate adaptation policies for urban regions
to the design, build and even manufacture
of high-performance environments, where
sustainable living and thriving commerce
are the norm, not the exception for the few.

The EAG has a one-year time frame during

which it will report to the RIBA Council each
quarter, with its final report and presentations
being made available by December 2024.

Philip D. Allsopp D.Arch., M.S.(Public Health),

RIBA, CSBA

RIBA © 2024 All rights reserved. No part of this report may be reproduced
or shared in any form or by any means, electronic or mechanical, including
photocopying, recording, or by any information storage or retrieval system,
without permission in writing from the copyright holder.

The content of articles contributed by external authors and published in

this report are the views of those authors and do not represent the position
of the Royal Institute of British Architects (RIBA).

Front Cover: Experiment with Midjourney, image courtesy of Jaina Valji, Copy and Space.

2
RIBA AI Report 2024

Foreword
We are currently in an era marked by rapid Within the architecture context, the choices we make about the
use of AI will shape the character of our cities, the quality of our built
technological advancement. Technology, environment and the well-being of future generations. In this context,
for me, is a better way of doing things and the Royal Institute of British Architects (RIBA) plays a crucial role
in guiding the discourse surrounding AI in architecture. By fostering
that can involve tools or machines. Today, interdisciplinary collaboration and promoting dialogue among
artificial intelligence (AI) emerges as the architects, technologists, policymakers and the public, we present
a pioneering exploration into the intersection of AI and architecture.
most disruptive tool of our time and its role Helping us to collectively chart a course through the complexities
in shaping the future of architecture cannot of AI integration while championing ethical principles and
human-centred design.
be overstated.
Through a series of expert articles and case studies, we examine
As we stand at the intersection of innovation and tradition, the the ways in which AI is already reshaping architectural practice, from
decisions we make about the integration of AI into architectural computational design and digital fabrication to urban planning and
practice will have profound implications for the trajectory of environmental sustainability. We also confront the ethical dilemmas
our profession and the built environment. As, in today’s rapidly inherent in the adoption of AI, exploring issues of equity, transparency
evolving landscape, technological innovation stands as and accountability in architectural decision-making.
a cornerstone of progress.
As we confront these challenges, it is essential that we approach
In this report, we embark on a comprehensive exploration of AI with a critical yet optimistic mindset, recognising its potential to
AI’s impact on architecture, navigating the complex landscape unlock new possibilities for innovation and creativity in architecture.
of possibilities and challenges that lie ahead. Just as Yuval Noah By fostering interdisciplinary collaboration and embracing a culture
Harari argues the direction of human history hinges on geopolitical of responsible innovation, we can harness the power of AI to create
events, the trajectory of architecture is significantly influenced more inclusive, resilient and sustainable built environments for all.
by advancements in technology.
Join us on this journey as we navigate AI in architecture and
At the heart of our inquiry lies a fundamental question: how can chart a course towards a future that is technologically advanced,
we harness the power of AI to enhance the practice of architecture yet ethically grounded.
while safeguarding the values that define our profession?
As architects, we are acutely aware of the transformative potential
of technological advancement, yet we also recognise the imperative
of responsible stewardship in the face of rapid change.

Muyiwa Oki
RIBA President 2023 - 2025

3
RIBA AI Report 2024

Digital twin technologies: a major opportunity

A digital twin of an urban region is an electronic model comprising The Adam Smith problem
several classes of analytic, simulation and visualisation technologies For over 250 years, commerce and industry have marched to the beat
that use, and indeed generate, numeric, descriptive and 3D data to of Adam Smith’s seminal work The Wealth of Nations, which viewed
test-drive policies and design criteria. In the fields of urban policy the earth’s resources as infinite and held the marketplace, and those
and design, digital twins can be used to simulate how, optimally, people whose interest in profits for personal gain superseded all other
to achieve a wide range of social, economic, mobility, energy, considerations, to be the drivers of competition and, according to his
water-use and other climate-related environmental outcomes. definition, innovation. By the end of the Second World War, Smith’s
The potential for transforming the guesswork involved in model of human behaviour and commerce was dominant and, by
policymaking is significant. 2020, had become supercharged to form a philosophy of planetary
ownership by corporate entities who, to this day, exert enormous
Rapidly advancing computational power built around advanced
influence over national and global politics in their headlong extraction
reduced instruction set computing machines (ARM for short),
of profits by whatever means possible.
pioneered 39 years ago in the UK, and emerging powerful AI
systems, such as ChatGPT or Adobe Firefly, greatly expand our It now looks like our species and thousands of others will be paying
ability to imagine the future and account for the massive complexities a very heavy price for the folly of our belief that the Earth’s resources
that exist in any place on earth where human beings congregate were infinite, despite the finite size of the planet, and that economic
and establish permanent settlements – the cities and urban regions ‘externalities’ created by commerce and industry would have
we all know so well. The architecture profession is generally familiar no discernable impact on lives, the price of goods or the planet.
with many of these technologies, such as building information There are few urban regions which have not been shaped by
modelling (BIM) systems, geographic information systems (GIS) these pernicious forces. Despite monumental efforts by scientists,
and other modelling software, but emerging technologies, including professionals and policymakers everywhere for better living conditions
advances in the field of system dynamics,1 3D visualisation and and better-performing built environments, the best-laid plans for
animation systems and synthetic data from curated AI applications, greater livability and healthier environments, social equity and
offer the potential for architects to take on far broader and more economic diversity continue to be based on the hope that private
impactful roles than ever before.2 sector development will come up with the answers. For decades
these ‘answers’ have been geared towards the extraction of profits
These profound opportunities for the architecture profession
to the exclusion of almost all other considerations. Hope faces
arise from a confluence of economic, legislative, environmental
powerful incentives to keep doing more of the things which we
and technological forces, which are making our urban regions
know are destroying our world. We have a choice to make, as a
ever more challenging for sustained livability to be the norm
species and as a profession – continue the march to oblivion based
for everyone, especially given the catastrophic effects of climate
on the concept of planetary ownership or shift everything about our
change happening globally.
society and what we do to the point where planetary stewardship
guides our actions (see Figure 1).

1
Jay Forrester, Urban Dynamics, MIT Press, Cambridge MA, 1969 and World Dynamics, Wright-Allen Press, 2
We have already seen the stellar results achieved by architects engaged in developing (coding and
Cambridge MA, 1971 (for more information on Forrester, see https://mitsloan.mit.edu/ideas-made-to- programming) and applying early forms of these technologies from the 1970s to the present day
matter/professor-emeritus-jay-w-forrester-digital-computing-and-system-dynamics-pioneer-dies-98). (such as Applied Research of Cambridge and Skidmore, Owings and Merrill, among many others).
One such example is David Rutten, an Austrian architect who devised and coded Grasshopper –
one of the most powerful parametric and generative design environments available today.

Philip D. Allsopp
RIBA EAG: Data, Computational Design
& AI, CEO of ORBIS Dynamics

4
RIBA AI Report 2024

Figure 1: Planetary Ownership or Planetary Stewardship?

Given politicians’ inaction over the past several decades, it might As soon as the engine is in service, its performance is monitored
appear that all is lost and we will all be drawn into an extinction by on-board instruments and, whenever a problem arises, a product
vortex from which escape is impossible. Although such an eventuality maintenance digital twin is also present to determine what improvements
is possible, it is by no means inevitable if we apply our collective or repairs need to be made to keep the engine safely in service.
know-how and technologies to adapt to climate realities and, at the
same time, slow the climate change flywheel which we have helped This does not happen in urban regions. As a society, we leave
to spool up over the past couple of centuries of intensifying industrial the design, location and performance of human habitats (the built
development. It is in the urban regions, where hundreds of millions environments which give form to and enable – or disable – a dizzying
of people live, that we have the best chance of making a difference array of human endeavours) to the whims of real estate speculation,
and changing the forces that currently pose such grave threats land-banking or to prescriptive ‘urban planning’ codes, many of
to life on Earth. which are devoid of any evidence basis and are often decades out
of sync with present and future human needs. The results to date
Today, the global construction sector lags behind all other sectors do not give much encouragement that doing more of the same
in innovation and productivity. Compared to most other industrial and will ensure better outcomes.
service sectors, whose productivity and performance have increased
by roughly 1,500% over the past 70 years, the construction sector While some components of digital twin technologies are indeed
barely breaks a 1% improvement.3 This presents the architecture used at the front end of the project design (including all construction
profession with a massive opportunity to leverage digital twin documentation and engineering calculations), they tend to operate like
technologies to move from being drafting services for developers a single bookend to a very full but open-ended shelf of books. What
to premier league players operating at more strategic, policymaking gets built is rarely, if ever, subjected to ongoing performance analytics
levels, where profound decisions are regularly made which shape to determine whether what was thought to be (and indeed simulated
the performance and livability of every urban area on the planet. as) a good solution actually worked when it came into contact with
people and the realities of their lives or the work they do. It is now very
apparent that new financial incentives are needed to ensure that the
Simulating the future by understanding urban DNA story of an urban region is bookended from design to in-use evaluation
If we were designing a jet engine for manufacture, we would build and management. Many other fields already do this – medicine,
a digital twin comprising 100% of its parts. We would then simulate consumer electronic products, aerospace, the automotive industry.
its performance against design goals, such as fuel efficiency, number Built environments remain by and large experimental prototypes that
of hours between maintenance, material degradation, integrity never go into production (paraphrasing Jony Ive, Apple’s former chief
following bird strikes and so on. When the achievement of those design officer). AI, data, generative design systems, system dynamics
criteria had been optimised according to the priorities given to each models and curated AI all can play powerful roles in shifting the status
one, the resulting jet engine’s digital twin would alter its form and quo to a better way of ensuring greater performance, better durability
shape to enable those complex sets of goals to be optimally achieved. and better human outcomes from what is built.
From that optimised digital twin, the manufacturing process starts.

3
Reinventing Construction: A Route to Higher Productivity, McKinsey Global Institute, February 2017 (https://www.mckinsey.com/capabilities/operations/our-insights/reinventing-construction-through-a-productivity-revolution).

5
RIBA AI Report 2024

Figure 2: An illustrative example of an urban region’s DNA components

Digital twin technologies enable architects to account for all the The future before us
genomic complexities of an urban region’s DNA that represent how AI, generative design systems, data science and the field of system
and why an urban region works in the way that it does. This offers dynamics make it entirely possible for architects to take leadership
significant opportunities for architects to participate in shaping the roles in urban policy deliberations rather than being left out of the
structure and form of urban regions of any size and location, driven picture until a developer has been given permission to build something
by urban genomic goals rather than hope, whimsy or artistic style-du-jour. within prescribed zoning rules. Such leadership roles are attainable if
There is no shortage of urgency for this endeavour. This reshaping of architects engage in the creation of digital models of cities that go well
urban forms needs to be done on both a large scale and an individual beyond BIM systems and reach into the human, social, economic,
building or urban site scale, including retrofitting what has already been mobility and environmental dynamics that drive sustainable livability,
built. The performance parameters for urban physical infrastructure better health outcomes, economic opportunity and prosperity.
are definable and we already know a great deal about them (see Figure 2).
In the field of medicine, such diagnostic technologies, involving AI
What if the form of redeveloped urban regions was driven by these and digital twins of the human body and its complex systems, are
parameters rather than purely by profit extraction and code compliance, used daily to detect diseases, conduct surgical procedures, apply
as so many ‘developments’ are today?4 The technologies underpinning gene therapies and benchmark progress towards recovered health.
digital twins are capable of identifying a variety of performance-defined In the field of built environments, the application of similar digital
urban ‘sand box’ locations. Within these, developers, owners, architects twin technologies opens up a breathtaking spectrum of possibilities
and allied professionals can put their know-how and creativity to good for improving the lives of billions of our fellow human beings and
use, knowing that what they built would have a higher probability of the planet we depend on for life.
enabling performance goals, like walkability, high air quality, mobility
choices and diverse local commerce, to be met rather than simply
hoping that they’d turn out OK in the end.

4
Judyta Cichocka, ‘Generative design optimization in urban planning: Walkability-optimized city concept’, Architectus, vol. 1, no. 41, 2015.

6
RIBA AI Report 2024

The current state, promise and near-term future

of AI in architecture
When designing our homes, workplaces, schools and communities, AI: empowering the new workforce of architects
the architecture industry is under constantly increasing pressures At the turn of this century, architects embraced 3D parametric
to help combat climate change, accommodate growing populations modelling through building information modelling (BIM), then
and meet stricter standards and requirements. To address these connected BIM in the cloud to improve coordination and productivity
challenges, the industry is undergoing a technological shift – in – bringing in an era of greater creative and technical collaboration.
addition to embracing 3D modelling, practices are adopting a more Now, AI promises to disrupt the practice (for the better) once again.
outcome-based way of working, one that is driven by AI, automation
and data-supported software applications. Although BIM revolutionised the industry, the issue of silos remains –
currently, BIM struggles to meet expectations to unify data and
Over the past two decades, new tools have been transforming the workflows across the entire plan, design, build and operate lifecycle.
way that architects work – from sketching concepts on a page to Integrating granular data, automaton and AI into existing workflows,
creating drawings in AutoCAD and, more recently, collaborating on while leveraging the design tools that architects use today, will create
building information models in the cloud. Architects have always been a more connected and outcome-based approach.
trailblazers in digital transformation, combining human creativity with
tech innovation. Now, AI opens up a myriad of new opportunities for Insights derived from data throughout the design process can simplify
automating routine tasks, empowering architects to solve even more everything from the exploration of design concepts, to evaluating
complex design challenges, such as delivering more sustainable environmental qualities surrounding a building site or how a building
building outcomes. performs in the real world.

Leveraging the ability of AI to augment, automate and analyse

gives architects their time back – not just by increasing productivity,
but by giving them the space to be more ambitious and to focus on
creative solutions. When used correctly, AI can help architects analyse
a myriad of design variations in a very short time, offering them
new perspectives on how to achieve important project outcomes.
This expands the realm of design possibilities, bringing us closer to
meeting increasing demands from building owners, developers, citizens
and municipalities, all while creating something sustainable, something
that can stand the test of time.

Project Phoenix in West Oakland, CA, showcases generative design, innovative materials like mycelium insulation, and carbon-neutrality. Image courtesy of Autodesk.

Amy Bunszel
Executive Vice President,
AEC Solutions at Autodesk

7
RIBA AI Report 2024

The AI analysis capabilities of Autodesk Forma show the impact of different design concepts on livability and environmental performance in real-time. Image courtesy of Autodesk

Sustainability and climate change resilience: intelligent insights in early-stage design

Societal challenges, such as rapid urbanisation combined with Take Project Phoenix, for example, a 316-unit modular housing
population growth and climate change, are putting pressure on the development in West Oakland, California, on a site that is heavily
architecture, engineering, construction and operations AECO industry. impacted by congestion and noise pollution. In a collaborative effort
The complexity of urban areas has also increased massively, and between MBH Architects, Factory OS and Autodesk, a multidisciplinary
climate change is fundamentally affecting the way people live and work. team harnessed the power of technology to share data and workflows
and tap into AI-powered insights across the project lifecycle to make
At the same time, the economic demands on the AECO industry are housing that was faster to design and build and more sustainable.
also increasing. The creativity of architects is being tested by the need
to maximise building density and use of space without negatively In the early phases of the project, the team made data-informed
impacting people’s quality of life or the environment. This is precisely trade-offs between goals for operational carbon, embodied carbon, cost
where AI and outcome-based design come into play, empowering and liveability. The team also leveraged their unit catalogue from past
architects to solve problems and enhance their current ways of working. projects to reduce time and risk through reusable design intelligence.
The shift towards outcome-based design, powered by AI, enables
architects to arrive at solutions faster and more efficiently. Targeting carbon neutrality, the housing units feature innovative
materials, including facade panels made from a core of mycelium, the
AI also makes it easy for architects to incorporate environmental rootlike structure of fungi. The panels themselves are carbon-negative,
and other contextual data into plans when optimising designs. as the amount of carbon captured in the mycelium core exceeds the
AI not only provides real-time analytics, that fuel essential insights carbon emitted by the process of making the panels.
into operational energy, microclimate, sunlight, wind and noise, it
also enables architects to test a wide variety of scenarios digitally, In the final stages of development, the team combined physical and
in a risk-free environment, to find optimal solutions within digital automation to construct a set of buildings that is not only efficient,
chosen parameters. but is also loved by residents.

8
RIBA AI Report 2024

The Phoenix will be built at about half the cost, time, and carbon footprint of a typical multi-family building in the San Francisco Bay Area. Image courtesy of Autodesk.

AI assistants: human expertise remains irreplaceable

Despite the benefits that AI brings to the architectural profession Combining human intuition and expertise with AI’s computational
today, it is understandable that many architects are still wary of the capabilities allows architects to explore more possibilities for
threat of job displacement. We envision AI serving as an assistant sustainable and innovative designs, leading to better informed
in the design process, with designers retaining their role as and more creative solutions. The integration of AI-enabled capabilities
decision-makers, controlling the creative process and ultimately into the design process does not mean architects will no longer
making the final call. It is the architect who has the real-world be required; instead, it empowers architects to focus on outcomes.
understanding of local specifics and needs – be they cultural and
aesthetic concerns, regulatory issues, such as local and regional So, what is next for the architect’s toolbox? It is an exciting time:
building codes, or the complex web of multi-layered relationships it is undeniable that AI is here to stay – both in our personal and
with stakeholders and customers. There is no doubt that machines professional lives. The practice of architecture is poised to leverage
can help with the heavy lifting, but it is humans who will be AI as an indispensable tool for transformation.
answering the all-important question of how to create better
As technology streamlines mundane processes and enhances
homes and buildings.
workflows, architects will gain more time to design solutions to
some of the world’s most pressing problems. In the face of rapid
change, one constant is the intuition and expertise of architects.
Architects will remain irreplaceable – essential for balancing
technological advancements with their understanding of human
needs and cultural values – in shaping the cities of tomorrow.
The humble pen and pencil will remain in the architect’s toolbox –
but now they are set up to work side-by-side with sophisticated
AI-supported digital tools.
The integration of AI-enabled
capabilities into the design process
does not mean architects will no
longer be required; instead, it empowers
architects to focus on outcomes.
9
RIBA AI Report 2024

Architects and artificial intelligence

Architects have been applying AI for discrete aspects of their With the pressing need to collate, learn from and disseminate this
projects for some time, working parametrically and generatively knowledge, and to deliver consistent and coherent solutions quickly,
using scripts and models, but in 2023 the topic went mainstream. AI offers the opportunity to collect new knowledge rapidly and to
Throughout that year, seemingly every article on technological disseminate it differently. It can place the most up-to-date learning
innovation focused on AI, tumbling from the roll-out of, and and reliable and robust data at the fingertips of the project team,
to a certain extent the boardroom antics behind, ChatGPT. be that the embodied carbon of a product or the plant space required
for a ground source heat pump installation.
This year, 2024, we are not likely to see any let-up in the conversation.
From a practice perspective, many have already put in place measures In this shift away from heuristic ways of working, the fact remains
and guidelines to ensure that their data is not compromised, and it will that significant judgment calls will still need to be made. The human
be important that ethics, cyber and data security and related topics is not disappearing any time soon, although by augmenting current
continue to be addressed to bring clarity to how AI tools are used. knowledge and skills with AI-facilitated knowledge, new value
propositions will emerge. This new partnership will result in unknown
However, there is an imperative to take a step back. AI is moving and unintended consequences, and to deal with these the built
from a narrow use in specific cases to an all-consuming application environment industry must engage with a wide range of new
across every aspect of projects, yet the reality for most is that AI professionals. Data scientists are an obvious start, but human factors
is an unknown entity and those implementing projects just want experts can look at the crucial interfaces between expert systems and
to know what it is and how they might use it. In this context, this humans. Anthropologists can help shape the processes of the future.
article outlines the strategic implementation of AI and how it might Decision-making will be transformed, allowing the project team to
be harnessed to create better outcomes for clients and a new spend more time on the challenges of the future, such as creating
generation of low-carbon, world-class buildings. materials that sequester carbon.
AI will benefit projects in various ways, making the day-to-day Of course, to get there requires carefully orchestrated steps, including
management of projects easier. Over recent years, communication the mundane task of properly classifying large amounts of metadata.
has spiralled out of control, as old forms of correspondence, such Evidence-based design is not a new idea, but AI, and the onset of
as emails, reports and file-sharing, mesh with new ones, including millions of data points, offers the opportunity to take this methodology
instant messaging and online meetings. Digital has made the to the next level. Finally, and most importantly of all, the future will not
management of projects more difficult, not easier, but AI offers be an evolution of past processes, but a revolutionary way of doing
the perfect solution to this challenge. things and such a paradigm shift is now inevitable. The key is preparing
today for this disruptive new way of designing and making buildings.
However, to unlock the true value of AI we must go further.
Transformation will not be achieved by optimising the traditional
ways of doing things, or by powering up building information modelling
(BIM), but by changing the essence of the design process itself.
To achieve this, we need to rethink the concepts at the heart of the
process. The timing is perfect. In the race to reduce carbon emissions,
many new solutions and systems are being deployed and the lead
designer can no longer rely on the inbuilt knowledge collected through
years of learned experience. Throw into the mix the shift towards
offsite manufacturing and it must be acknowledged that the design
process has simply become too complex, and the knowledge
required to create a new generation of buildings too vast.

Dale Sinclair
Head of Digital Innovation at WSP

10
 RIBA AI Report 2024

Machine Learning: Architecture

in the age of Artificial Intelligence

The architecture profession is changing. Practices must stay abreast of

new developments in AI or risk being left behind. Architecture’s best-known
technologist, Phil Bernstein, provides a strategy for long-term success.

Buy Machine Learning: Architecture

in the
TheAge of Artificial
advent of machineIntelligence MACHINE ARCHITECTURE IN THE AGE OF
learning-based AI systems LEARNING ARTIFICIAL INTELLIGENCE
online, or stop by the bookshop at
demands that our industry
tion
66 does
Portland Place,share
not just London.
toys,
pal but builds a new sandbox
n
in which to play with them.
In)
at
The profession is changing. A new era is rapidly
approaching when artificial intelligence will augment
ons

‘The advent of machine

the work of architects, making the design process faster,
better coordinated, more accurate and rooted in data.
The danger, however, is that, without a clear strategy

learning-based AI
to direct new technologies, they will encroach on the
difficult and ambiguous work of architects – to the
detriment of the profession and the built environment.

systems demands that

Leading architectural technologist Phil Bernstein provides
that strategy. Divided into three key sections – Process,
Relationships and Results – Machine Learning lays out an

our industry does

approach for anticipating, understanding and managing
a world in which computers often augment, but may well
supplant, knowledge workers like architects. Armed with

not just share toys,

this insight, the profession can take full advantage of
the new technologies to future-proof its business.

but builds a new

Features chapters on:

Professionalism

sandbox in which to
Tools and technologies
Laws, policy and risk
Delivery, means and methods
Creating, consuming and curating data

play with them.’

Value propositions and business models.

Phil Bernstein ISBN 97819-1-412-401-3

9 781914 124013

Follow us:
RIBAPublishing
RIBABooks
1
RIBABooks.com
RIBA Xxx 11
RIBA AI Report 2024

Experiment 2 with Midjourney, image courtesy of Jaina Valji, Copy and Space

RIBA AI survey: findings

Introduction
As new tech tools became available over the past year to 18 months, there has been an
increased intensity to the discussion about the future of artificial intelligence (AI) and what
it means for society generally and for the architectural profession specifically. Too often,
opinion seems irrevocably divided, between existential despair (the end of architects, the
end of professions, the end of work) and a feeling of the untouchable primacy of human
creativity (no machine could create as we create).
This discussion has not, however, been supported by evidence, AI promises a beneficial and rapid evolution of the architectural
as there has been very little information available about what is profession; routine tasks automated, the design process streamlined,
happening in practice. This is where the RIBA AI Survey aims to help, carbon reduced, creativity accelerated and expanded, and client
by gathering information about the current and near-term realities of and societal outcomes improved. This survey looks at how far that
AI in architecture. The findings suggest that, while a significant number promise might be realised now and within the next two years.
of practices have started to use AI in at least some of their projects
(41%), AI adoption in the profession is in its infancy, with many There are risks too, including further pressure on fees and job
practices not using AI or using it only occasionally. However, adoption displacement, and these are also explored.
and use of AI are set to increase in the coming years.

Adrian Malleson
Head of Economic Research and Analysis
Royal Institute of British Architects

12
RIBA AI Report 2024

Architecture and digital maturity

Digital innovation in architecture is ongoing and AI is its latest Digital maturity
manifestation. Respondents were asked to rate their organisation’s digital maturity,
and the responses suggest a well-distributed range.
Architects have taken a leading role in the digitisation of the
construction sector, exemplified by the professions’ pioneering A small but notable 11% consider themselves as leading digital
adoption and use of building information modelling (BIM). innovators, with a further 19% identifying themselves as early adopters
This makes sense. If the profession does not continue to adopt, of digital innovation. This suggests around three in ten practices actively
adapt to and lead new technologies, it might not only fail to reap look to develop their practice offering through leading in digitisation.
the benefits of innovation but also leave unguarded significant
areas of current and future work and, therefore, revenue. A near-majority (47%) believe their digital maturity is around where
most are, suggesting a willingness to adopt new digital tools that are
already of proven value.

At the other end of the range, 17% are late adopters of digital innovation,
and 5% resist digital innovation altogether, preferring traditional
techniques. This resistance might be due to a lack of resources or
in-house skills. Alternatively, it might be that for practices that take
on particular types of work, digitisation has a more limited role.

Overall, how would you assess your organisation’s digital maturity?

5% 17% 47% 19% 11%

We tend to resist digital We tend to be late Our digital maturity We tend to be early We think of ourselves as
innovation, preferring more adopters of digital is around where adopters of digital being among the leading
traditional techniques innovation most organisations in innovation digital innovators
our sector are

A near-majority (47%) believe their

digital maturity is around where most are,
suggesting a willingness to adopt new digital
tools that are already of proven value.
13
RIBA AI Report 2024

During the work stages for which your practice is commissioned, do you create and maintain building models
in accordance with ISO 19650?

Never 34%

Sometimes 28%
Always 26%

Rarely 12%

Before AI: BIM

For most of this century, digital maturity has been associated with BIM. Survey findings suggest there is a range in the extent to which
The architectural profession has successfully adopted BIM into its practices create and maintain building models that comply with
workflows and as a result is now well placed to implement AI. ISO 19650. A quarter of the respondents (26%) always create
and maintain models that comply with ISO19650, a slightly higher
The bedrock of BIM is well-structured data. AI cannot get off proportion (28%) sometimes do, while 12% rarely do and just over
the ground without good training data – which BIM can provide. a third (34%) never do.
So, before examining AI, the survey looked to uncover the extent
to which practices work with well-formed model data. Small practices are significantly less likely to create and maintain
ISO 19650 compliant models, and large practices are significantly
ISO 19650 1 is a series of international standards that describes more likely to. Twelve per cent of small practices with between one
a collaborative and consistent approach to information management and ten staff always create and maintain BIM models that comply
for built assets. It forms part of the UK BIM Framework. The series with ISO 19650 and 62% never do. This ‘always’ figure rises to
includes concepts and principles (ISO 19650-1), the asset delivery 43% for large (50 to 99 staff) practices, and 50% for those
phase (ISO 19650-2), the asset operational phase (ISO 19650-3), practices with 100 or more staff.
information exchange (ISO 19650-4) and information security
(ISO 19650-5). There is a resource of well-structured data in many but not all
practices. This may prove to be a foundation for the future
development of AI within the profession.

1
https://www.bsigroup.com/en-GB/capabilities/buildings-and-construction/iso-19650-building-information-modelling-bim/

14
RIBA AI Report 2024

Knowledge and current use of AI

Knowledge of AI
AI adoption is in its infancy, with just 2% of practices using it on Almost all architects have at least some knowledge of AI.
every project. But two-fifths use it on at least the occasional project. The majority (51%) assess themselves as having a basic knowledge,
and a third (32%) as having a practical knowledge of AI. A small
Personal knowledge about AI is running ahead of its adoption by percentage have advanced knowledge (6%) or are recognised
organisations. authorities (2%). In contrast, fewer than one in ten (9%) architects
have no knowledge of AI.

Overall, how would you rate your personal knowledge about AI, in general?

9% 51% 32% 6% 2%
No Basic Practical Advanced Recognised
knowledge knowledge knowledge knowledge authority

Use of AI
AI adoption in practice stands at 41%. Significant numbers of practices Most commonly, practices that have adopted AI use it for some
are using AI for at least the occasional project. That said, the use of AI projects (15% of all) or the occasional project (20%). Should AI
in architectural projects is not the norm, with 59% of practices never become the norm, we are at the start of the adoption curve.
using AI, only 2% using it for every project, and just 4% using it for
most projects.

For the projects you are currently working on, how often does your practice use AI in any way?

2% 4% 15% 20% 59%

For every For most For some For occasional
project projects projects projects Never

AI adoption in practice stands at 41%.

Significant numbers of practices are using
AI for at least the occasional project.
15
RIBA AI Report 2024

The profession’s views on AI

Architects’ views about AI vary, and not all are positive. As well as However, AI also has a transformative potential. Modern buildings
seeing the potential benefits, the profession also sees significant risks. are becoming increasingly complex in their design, construction and
A common view is that AI is a threat to the profession; that it will maintenance. A near majority (49%) of architects agree that this
destroy jobs by automating some or all of the roles architects fulfil. complexity means the profession needs more and better digital tools,
Architects are evenly split on this possible future, with 36% agreeing including AI. Practices could gain a competitive edge here by being
that AI is a threat to the architectural profession, 34% disagreeing and early AI adopters, but this potential is likely to need investment to
30% neither agreeing nor disagreeing. A future without the profession realise. Just one in five (20%) practices have invested resources in
is being seriously considered by some, suggesting architects need AI research and development. The majority (69%) have not.
to engage with and shape the future of AI now.
There is a balance that needs to be worked out, between making the
Does AI present a risk of imitation? A majority of architects (58%) most of (and investing in) the transformative potential AI offers, and
think it does. Architecture is a creative profession, with creativity safeguarding creativity and associated intellectual property. This is not
expressed through building design. If AI can easily and cheaply output just an issue for architects, but one for the wider creative industries.
plausible design imitations that can be readily passed-off, the creative
foundation of the profession may become vulnerable.

Please state how strongly you agree or disagree with the following statements about AI now:

AI increases the risk of our work being imitated

58%
24%
18%

Building design is so complex now, we need more and better digital tools, like AI

49%
29%
22%

AI is a threat to the profession

36%
30%
34%

My practice has invested resources in AI research and development

20%
12%
69%

Agree Neither Agree nor Disagree Disagree

16
RIBA AI Report 2024

Views of current AI users

Forty-one per cent of respondents are using AI in their projects, but 59% are not. This section looks at the views of that 41%, those who are
currently using AI in some way.

Current integration and benefits of AI

Of those using AI in some way, 43% agree that AI has improved Only 24% agree they have successfully integrated AI into bid
efficiency in the architectural design processes, while 24% disagree. creation, project management or scheduling, while a substantial
This is the one area where current users of AI, on balance, see a 48% disagree. Twenty-one per cent agree that they have employed
clear benefit. AI in environmental sustainability analysis, while 46% disagree.
This level of integration is set to rise in the near term.
The benefits are less clear in the other areas asked about. More
respondents disagree (39%) that AI currently enhances accuracy Just 7% agree that AI has led to staff reductions and 61% disagree,
in modelling and simulations than agree (26%). Only 11% agree suggesting that AI adoption is not significantly affecting practice
that AI enhances the accuracy of specifications, while a majority staff levels. While the most apocalyptic of AI writers describe the
(49%) disagree. professions as next in line for employment decimation or obliteration,
of the kind we saw for land workers during the agricultural revolution
This lack of perceived benefit might be because current AI tools or for craft workers during the Industrial Revolution, there is little
are not yet sufficiently developed, or because most architects lack evidence of this currently being widespread. Views about the possible
the training and guidance needed to make the most of them. future are explored more later.
Or perhaps this is typical in the early stages of the adoption of
a new technology. The benefits to practice are expected to increase
over the next two years.

For those using AI - How strongly do you agree or disagree with the following statements about AI:

AI has improved efficiency in our architectural design processes 43% 33% 24%

AI has enhanced the accuracy of our architectural modelling 26% 35% 39%
and simulations

AI has been integrated into our bid creation, project management, 24% 28% 48%
or scheduling

AI has been employed in our environmental sustainability analysis 21% 33% 46%
(e.g., energy efficiency, material optimisation)

AI has enhanced the accuracy of our specifications 11% 40% 49%

AI has led to staff reductions 7% 32% 61%

Agree Neither agree nor disagree Disagree

17
RIBA AI Report 2024

AI and the design process

Among respondents who are using AI in some way, the most common Forty-three per cent never use AI for model generation, while only
use is for early design stage visualisations, with 6% using it always, 2% always do, with 12% using AI often here, 22% sometimes and
22% often, 60% sometimes or rarely, and only 12% never using it 21% rarely.
for this purpose. This use of AI may, for example, help clients to see
the possible resolutions of their brief more clearly, through detailed, An overwhelming 61% do not employ AI for specification writing.
even immersive, visualisations. Creating the specification is possibly one of the least-loved parts
of the design process and it might offer a significant opportunity
Twenty-one per cent use AI for generative design always or often, for automation through AI, which could improve accuracy, material
while 31% use it sometimes, 16% rarely and almost a third (32%) and product choice, and consistency with other sources of
never. AI-based generative design has the potential to fundamentally design information.
change the design process, allowing architects to rapidly create new
and innovative designs, which can be explored, analysed and then Only a minority of those who use AI use it for construction product
refined to meet the brief in new and better ways. and material selection and analysis, building performance simulation,
standards and regulatory compliance checking or environmental
Parametric design has been a common feature of design tools impact modelling. There my be significant potential here. To take
since the early days of BIM and 3D design. AI has the potential one example, an AI tool that can deliver accurate and speedy
to optimise and extend the use of parametric design, by drawing regulation and compliance checking could quicken planning
on wider and more complete datasets, allowing workable choices application progress while helping to ensure that buildings are
of building elements and systems to be algorithmically generated. safe, accessible and sustainable.
Forty-three per cent of those using AI in some way always, often
or sometimes use it for parametric design, while 17% rarely and
40% never use it for this purpose.

Please indicate how far AI has been adopted within your organisation in the following areas of the design process:

Early design stage visualisations 6% 22% 42% 18% 12%

Generative design 4 17% 31% 16% 32%

Parametric design 2 14% 27% 17% 40%

Model generation 2 12% 22% 21% 43%

Building performance simulation 3 10% 14% 18% 54%

Environmental impact modelling 3 8% 14% 14% 60%

Specification writing 8% 13% 18% 61%

Standards and regulatory compliance checking 14 22% 10% 63%

Construction product and material selection and analysis 13 22% 18% 56%

Always Often Sometimes Rarely Never

18
RIBA AI Report 2024

AI and project management

The graph below shows the use of AI in project management. In all other areas of project management covered in the survey:
Generally, AI-using practices have been slower to adopt AI in project • project resource management
management than in the design process.
• cost information and modelling
Within project management, AI is only used significantly • project scheduling
in bid creation, where 10% have used AI sometimes or often, • fee calculation
19% sometimes and 25% rarely. That still leaves a significant
• project cost management and
46% who have never used AI for this purpose.
• contract management

a clear majority do not use AI, and only a small minority (10% or less)
use it often or always.

With the business of architecture under significant pressure to remain

profitable, there may be value in exploring the potential of AI to pick
up project administration, leaving architects free to develop client
relationships and create buildings.

Please indicate how far AI has been adopted within your organisation in the following areas of project management:

Bid creation 2 8% 19% 25% 42% 46%

Project resource management 2 7% 8% 14% 69%

Cost information and modelling 2 6% 13% 12% 68%

Project scheduling 1 5 12% 12% 70%

Fee calculation 15 11% 13% 71%

Project cost management 2 3 10% 11% 73%

Contract management 1 4 8% 11% 76%

Contract selection, editing and agreement 1 3 10% 11% 76%

Always Often Sometimes Rarely Never

19
RIBA AI Report 2024

AI – the near-term future

The survey also explored the near-term outlook for AI adoption A near majority expect AI to enhance accuracy in modelling and
and use. This set of questions asked respondents to consider what simulations (49%). A significant minority expect AI will be integrated
might happen over the next two years. The questions were put to into bid creation and project management (41%) and will come
all respondents, not just those using AI. The findings suggest that to enhance the accuracy of their specifications (40%).
AI will be widely adopted and integrated into practice and will
deliver tangible benefits in the near term. But there are risks too. However, many of the comments received expressed the view that role
of AI will always be limited. For example, that AI could never be
A majority of respondents (54%) agree that in two years’ time well-suited to considering the cultural, historical and social factors of
AI will have been adopted in their practices, although a quarter (25%) design, nor the intricate spatial, structural or regulatory complexities of
disagree. The remainder (21%) are equivocal. This anticipated adoption the design process, nor to make the subjective judgments of aesthetics
is not quite matched by investment, with 41% anticipating that their and client preference.
practice will invest in AI research and development.

A majority also agree that AI will be used to carry out environmental

sustainability analysis (57%) and that it will improve efficiency
in architectural design (57%).

Agreement that in the next 2 years:

AI will be employed in environmental sustainability analysis 57% 30% 13%

(e.g., energy efficiency, material optimisation)

AI will improve efficiency in our architectural design processes 57% 25% 18%

AI technologies will be adopted in my architectural practice

54% 21% 25%

AI will enhance the accuracy of our architectural modelling and simulations. 49% 28% 24%

My practice will invest resources in AI research and development

41% 24% 34%

AI will be integrated into our bid creation, project management,

or scheduling
41% 30% 30%

AI will enhance the accuracy of our specifications

40% 38% 23%

Agree Neither Agree nor Disagree Disagree

The findings suggest that AI will be widely

adopted and integrated into practice and
will deliver tangible benefits in the near term.
But there are risks too.
20
RIBA AI Report 2024

The anticipated increase in AI adoption is not without risk to the A majority do not see the oft-postulated existential risks to the
profession. While 47% agree that more digital tools, such as AI, will profession and employment coming in the next two years. But
be needed because of the increasing complexity of building design, significant numbers do. Thirty-six per cent agree that AI will lead
a majority (59%) agree that AI brings with it a risk of work being to staff reductions, while 30% disagree and 34% have no clear view.
imitated, perhaps as designs are appropriated by AI tools for use The view on the potential threat to the profession is finely balanced,
in AI training data, so becoming readily replicable in both spirit with 35% agreeing that AI is a near-term threat to the profession,
and detail. 36% disagreeing and 29% equivocal.

Agree

Neither Agree nor Disagree

23% 24%
Disagree
Building design
AI will increase will become so
the risk of our work complex, we will 47%
being imitated need more and
59% better digital
18% tools, like AI

29%

30%
36% 36% 35%
AI will lead AI will become
to staff a threat to the
reductions profession

34% 29%

A majority (59%) agree that AI brings with

it a risk of work being imitated, perhaps as
designs are appropriated by AI tools for use
in AI training data, so becoming readily
replicable in both spirit and detail.
21
RIBA AI Report 2024

Evaluation of AI
The construction industry and the design team face several Productivity and collaboration
long-standing challenges. From Latham 2 to Egan 3 to Farrell 4, the Sixty-five per cent of respondents think that AI will have a positive
construction industry has been characterised as adversarial, siloed, effect on the productivity of the construction industry, and only 10%
insufficiently collaborative and failing to make the productivity gains think it will have a negative effect. Half think that AI will have a positive
seen in other sectors. BIM may have helped, as a part of a programme effect on collaboration between architects and other professions, and
of wider digital transformation 5 of the sector, but perhaps much of the only 14% think the effect will be negative. On balance, the effect of
early promise remains unrealised. AI on project collaboration is anticipated to be positive, with 48%
expecting AI to improve project collaboration and just 13% feeling
AI might help address these challenges. Respondents were asked
the effect here will be negative.
whether AI would have a positive or negative effect in some important
areas. On balance, the effect of AI is seen as positive. Perhaps because architects collaborate well together already, 49%
think Al will make no difference. But even here, the balance is for
AI to have a positive effect, with 31% believing it will be positive for
collaboration between architects, and 20% that it will be negative.

Overall, do you think the effects of AI will be positive or negative?

Increasing the productivity of the construction industry 65% 25% 10%

Collaboration between architects and other professions 50% 36% 14%

Project collaboration
48% 40% 13%

Collaboration between architects 31% 49% 20%

Positive No Difference Negative

2
https://constructingexcellence.org.uk/constructing-the-team-the-latham-report/
3
https://constructingexcellence.org.uk/rethinking-construction-the-egan-report/
4
https://farrellreview.co.uk
5
https://www.mckinsey.com/capabilities/operations/our-insights/decoding-digital-transformation-in-construction

22
RIBA AI Report 2024

Carbon, innovation, creativity and education

With the built environment accounting for around 40% of greenhouse On balance, respondents suggest that AI will have a positive effect
gas emissions, design professionals and the wider construction on both design innovation (54% positive) and design creativity
industry have a responsibility to decarbonise construction. AI won’t (48% positive). This positivity is not universal, however, with the effect
solve the climate emergency (and is itself carbon intensive), but the seen as negative by 23% and 31% of respondents, respectively.
survey suggests this is an area in which AI can have a positive effect.
Sixty-four per cent of respondents believe AI, with its potential to Forty-four per cent of respondents anticipate AI having a positive
optimise energy usage, will contribute positively to meeting net-zero effect on architectural education, with AI having the potential to offer
targets. Aligned with this is the view that AI is well placed to help tailored on-demand learning and immediate feedback. The positive
create buildings that better meet performance requirements, outlook is not universal, however, with 36% anticipating a negative
with 63% anticipating that AI will have a positive effect here. effect. The experience of COVID-19 lockdowns reminds us of the
dangers 6 of digital-only education.

Overall, do you think the effects of AI will be positive or negative in?

Meeting net-zero targets 64% 29% 7%

Creating buildings that better meet performance requirements 63% 28% 9%

Design innovation
54% 23% 23%

Design creativity 48% 21% 31%

Architectural education
44% 20% 36%

Positive No Difference Negative

6
https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthandwellbeing/bulletins/coronavirusandthirdyearhighereducationstudentsengland/29novemberto20december2021

23
RIBA AI Report 2024

Risks remain
The most significant areas of concern are fees and employment. These issues, the impact of AI on employment, fees and the
Without adequate fees, there will be fewer jobs in architecture, future of the profession, are explored in more detail in the expert
and ultimately no profession. Only 15% of respondents think that articles featured in this report. To preserve and enhance their roles
AI will have a positive effect on fee income, and a clear majority and income, architects may need to reimagine what they do and how
(56%) believe the effect will be negative. A significant minority (46%) they charge for it. There are real opportunities, but these significant
anticipate negative effects on employment opportunities and only 22% risks must not be ignored.
see positive effects here.

Overall, do you think the effects of AI will be positive or negative?

15%
22%

46%

56% 30%
Employment Opportunities Increasing Professional
for Architects Fees
33%

Positive No Difference Negative

The most significant areas of concern are

fees and employment. Without adequate
fees, there will be fewer jobs in architecture,
and ultimately no profession.
24
RIBA AI Report 2024

Final word – ethical considerations

As a profession, architecture has a responsibility to practice within Ethical concerns around dealing with fellow members of the practice,
ethical boundaries. The RIBA Code of Professional Conduct7 fellow professionals, the wider design team and contractors were the
makes explicit the ethical requirements of RIBA membership, least pronounced; concerns relating to dealing with clients and the
and the RIBA Ethical Practice Guide8 and the Final Report9 of the wider community were the most pronounced.
RIBA Ethics and Sustainable Development Commission explore
and outline the profession’s ethical responsibilities in more detail. Adopting AI is not just a technological challenge, but an ethical one too,
particularly when considering those outside the design and construction
AI may bring with it ethical considerations, such as the risk of sectors – the clients and communities the profession seeks to serve.
inadvertent plagiarism, how to attribute ownership of and charge The findings suggest there is a need for the profession to reflect upon
a fair amount for work, and the changed relationship between and refine its responsibilities as the ethical implications of AI become
the architect and others where machine-generated design and clearer, to ensure continued trust and professional standing.
communication is interposed between them.

The survey results suggest architects tend to agree that AI brings

with it ethical considerations. For each of the architect relationships
given below, a clear majority of respondents suggest there are ethical
concerns, and a noteworthy proportion (21–27%) rank the ethical
concerns as ‘significant’.

Do you foresee ethical concerns arising out of the adoption of AI, in professional responsibilities towards:

Clients 27% 57% 16%

The wider community 27% 55% 18%

My fellow professionals
26% 49% 25%

Fellow members of my practice 23% 42% 35%

The wider design team

22% 47% 31%

Contractors
21% 43% 36%

Significant ethical concerns Some ethical concerns No ethical concerns

About the survey

The survey ran from October to November 2023, with The respondents were self-selecting, so the results are best read
RIBA members asked to share their views. Over 500 people as a very good indication of AI in the profession, but not as definitive.
responded to the survey; our sincere thanks to those who did.
Not everyone responded to every question (in part because RIBA would like to thank all those who took the time to complete
not every question was relevant to every respondent). the survey. It is greatly appreciated.

7
https://www.architecture.com/knowledge-and-resources/resources-landing-page/code-of-professional-conduct
8
https://www.ribabooks.com/RIBA-Ethical-Practice-Guide_9781914124723
9
https://www.architecture.com/knowledge-and-resources/resources-landing-page/ribas-ethics-and-sustainable-development-commission-final-report

25
RIBA Academy
HORIZONS 2034
WEBINAR SERIES
Which global megatrends The RIBA Horizons 2034 leadership webinar programme
will provide a ten-year view that imparts valuable insights
are set to shape society, into the near future. As the urgency of the climate crisis and
demographic pressures fully surface, alongside the far-
the built environment reaching impact of artificial intelligence (AI), engaging with
change has never been more pressing.
and the profession in the The programme includes:
decade to come? • Six webinars featuring expert voices including Dr Ronita
Bardhan, Stephanie Edwards, David Miller, Phil Bernstein and
more, beginning with the first module The Future Now, which
is free to attend.
• Unique insights into emerging challenges and opportunities
that can inform long-term decision-making, whether
personal career choices or design and business decisions.
• Focus on four core themes: The Environmental Challenge,
The Economics of the Built Environment, Population Change
and Technological Innovation; and how they intersect.

Find out more

and register
1 RIBA Xxx 26
RIBA AI Report 2024

Adoption of AI and its use within the

architecture industry
While the conversation surrounding AI is thriving within the AI in architecture and how is it used within the industry
architecture industry, is the actual adoption of AI by architects Let us begin by looking at what we mean by AI within the architecture
following the same path? The RIBA AI Survey sheds light on whether industry and how it is currently being used. AI within architecture
or not the industry is indeed turning to AI technologies to support refers to the application of computational techniques, algorithms
design processes and project workflows. and technologies to assist architects and designers at various stages
of the architectural process. It is already being utilised in various ways,
AI technologies have the potential to transform traditional
often without even being labelled as AI. Some of the more common
approaches to design and construction, offering architects powerful
areas of its current application are given below.
tools to enhance efficiency and creativity. However, there seems to be
significant hesitation among architects about whether to fully adopt Generative Design: AI algorithms generate and optimise design
and embrace AI. This appears to stem from uncertainties about how solutions based on criteria such as space requirements and aesthetic
AI will affect traditional roles within the industry, job displacement preferences. This approach allows architects to explore innovative
and both the reliability and the ethical implications of AI-driven options efficiently.
decision-making. These concerns are reflected in the results of
the RIBA AI Survey. For instance, a third of the architects surveyed Project management and scheduling: AI is being used in Customer
agree that AI will become a threat to the profession, and a third that Relationship Management (CRM) and cloud-based project
AI will lead to staff reductions. As a result of this uncertainty, many management software/systems to make predictions, optimise
architects are hesitating to fully embrace AI technologies, leading schedules, carry out risk assessments, and more.
to a relatively slow pace of adoption and integration within the field.
Building information modelling (BIM): AI is prevalent in existing BIM
practices. For example, AI algorithms can analyse BIM models and
provide valuable insights. It is also responsible for detecting clashes
between architectural, structural and mechanical systems, thereby
reducing errors and conflicts during construction.

Digital twins: Digital twins enable architects to simulate an existing

building at maximum accuracy. This can be achieved through point
cloud models of existing spaces, which can be used to create
BIM models, or by providing a collaborative platform for architects,
clients and stakeholders, where they can view and interact with
a virtual model.

Energy efficiency and sustainability: AI algorithms can optimise

building performance and predict energy use, daylighting, thermal
properties, and more.

Technology interactions: AI can help architects and engineers

to conduct structural analysis, analyse loads and optimise designs
for maximum strength and safety. There may be a happy time
when AI will produce building control drawings and specifications!

Jaina Valji
Architect and Founder of Copy and Space

27
RIBA AI Report 2024

Screenshot of a point cloud scan of Pitzanger Manor, image courtesy of Jaina Valji, Jaina Valji conducting a point cloud scan of Pitzanger Manor, image courtesy of Jaina Valji,
Copy and Space. Copy and Space.

AI and its implementation through BIM Challenges and risks of AI in the architecture industry
All of the applications outlined above can be implemented through With the great potential of AI comes inherent risks and challenges.
the use of BIM, which has a crucial role to play in the integration Some of the key challenges that AI may pose within the architecture
of AI within the architecture industry. BIM is a digital representation industry are examined below.
of the physical and functional characteristics of a building, providing
a collaborative platform where architects, engineers, contractors Overreliance on certain aspects of AI: Insufficient human oversight
and other stakeholders can design, visualise, simulate and manage could lead to unchecked biases or errors in AI-generated designs,
buildings throughout their life cycle. or to designs that unintentionally imitate copyrighted material.
To mitigate this risk, architects should remain critical of everything
When combined with AI technologies, BIM becomes even more that AI produces, leveraging AI as a tool to enhance, rather than
powerful, enabling enhanced automation and analysis and optimising replace, human expertise and creativity.
decision-making capabilities. By applying machine learning algorithms
to BIM data, AI can identify patterns, optimise design parameters and Legal and insurance considerations: Given the advancements of
predict project outcomes more accurately. The integration of AI with generative design iterations and technology interactions, it is important
BIM empowers architects to leverage the full potential of their digital for architects to remember that AI is not an entity that can be held
models. I believe that the use of BIM offers the greatest opportunity liable. Architects hold PI insurance and assume liability for all
for integrating AI within the architectural workflow – having a single information produced as a result of the use of AI.
collaborative model that allows repositories of information to be
Accuracy of the data used to train AI: The integration of AI in
accessed using AI.
architecture has several attendant risks relating to the data used to
train AI. For instance, biased, inaccurate or incomplete training data can
lead to designs that introduce societal biases, result in inaccuracies in
predictions and give rise to privacy concerns relating to sensitive data.
To mitigate these risks, a diverse and transparent process of data
selection should be applied with rigorous validation processes.

Even as an advocate of AI, I admit that achieving the balance is

undoubtedly tricky. On the one hand we, as architects, are asked to
embrace AI while, on the other hand, we are asked to be cautious of it
and double-check everything it produces. Despite AI’s capabilities,
there remains a need for human oversight and validation to ensure the
accuracy, reliability and ethical integrity of the outcomes it produces.
This dual responsibility requires architects to embrace AI as a valuable
The integration of AI with BIM tool while also maintaining an observant stance, double-checking and
verifying its outputs to mitigate potential errors or biases. If we take
empowers architects to leverage the this approach, we can ensure responsible AI adoption while driving
full potential of their digital models. innovation in architecture.

28
RIBA AI Report 2024

Experiment 3 with Midjourney, image courtesy of Jaina Valji, Copy and Space

Is AI going to steal our jobs? Conclusion

This is possibly one of the greatest perceived threats related to Fostering transparent communication between industry experts,
AI generally, including within the architecture industry, as shown by AI developers and architectural professionals will help to clarify the
the results of the RIBA AI Survey. I do not believe that AI will replace misconceptions surrounding AI and build trust in AI technologies.
the role of the architect but it will certainly reshape the nature Facilitating collaboration between architects, AI developers and other
of the profession. While AI technologies are increasingly capable stakeholders encourages a shared understanding of AI’s potential
of automating certain tasks within architectural practice, such as and its limitations. By working together on pilot projects and research
drafting, modelling and analysis, they are not capable of replacing initiatives, architects can gain hands-on experience of AI technologies
the creative and critical thinking abilities of human architects. and build confidence in their application within the industry.

Instead, AI is positioned to complement the capabilities of architects, Establishing clear ethical guidelines and regulatory frameworks
offering powerful tools for design exploration and optimisation. at a higher regulatory level to govern the use of AI in architecture
AI allows architects to tackle complex challenges in design and would help to alleviate concerns about potential misuse or
construction more efficiently and effectively. It facilitates faster unintended consequences. Architects need assurance that AI-driven
iteration, exploration of design alternatives and evaluation of decision-making processes both adhere to ethical standards and
performance criteria, leading to better-informed design decisions prioritise societal well-being.
and more innovative solutions.
Finally, integrating AI into architecture schools would not only
Ultimately, I believe that the role of architects in the era of AI is likely prepare students for the future of the profession but also teach
to evolve. Part of our role will be to learn how to successfully create them to leverage AI technologies responsibly and ethically to address
and implement parameters so we can leverage AI as a tool to enhance contemporary architectural challenges. Teaching students how
our creativity and expertise, thinking beyond the limits of the human to properly benefit from and safely use AI does not need to detract
mind and without time constraints. For example, if we wanted AI to from learning the fundamental skills of being an architect.
produce design options for 30m2 apartment arrangements that
were compliant with building regulations and London housing design Addressing challenges and fears within the industry through
standards while also maximising the use of daylight, we would need education and transparent communication about AI’s capabilities
to know how to ask AI to go about this task. Therefore, our roles will and limitations is critical to paving the way for more widespread
adapt to include learning how best to extract this information from the adoption of AI in architecture.
platform we are using in the same way that ChatGPT produces the
most useful information when you ask it a precise, specific question.

29
RIBA AI Report 2024

The Perceptron, invented by Frank Rosenblatt in 1958: The origins of deep-learning from machine perception. Image courtesy of National Museum of the US Navy

The computable and the incomputable

In the RIBA Journal of February 1967, Roger Walters described The quantified
design with computers as a series of choices or decisions involving If we keep to the historical narrative, AI is the latest form of
complex data: ‘At one end of the data spectrum were complete automation that aims to turbocharge the efficiency of a well-defined
quantifiable factors, at the other end the factors which could never task. Provided with a dataset, machine-learning will create a model –
be quantified, which derived from an architect’s creativity and turned capable of doing something, but learned through examples and
buildings into architecture.’ The debate then, as now, considered comparison rather than based on rules. It can replace certain
how computers might transform design and which tasks they machinic functions and absorb the quantifiable thinking. Generative
could and should be allowed to undertake. machine learning can create a letter, a sketch or an image faster.
Pre-qualification questionnaires or reports can be effortlessly improvised
In the 50 years since Walters wrote this, computing has not just
by ChatGPT with some simple prompting. The results may be relatively
delivered new drawing tools, but also fuelled a wider systemisation of
generic, and might sometimes contain inaccurate hallucinations, but
the workplace. The role of the architect is increasingly described through
through a little supervision we can be fairly clear on whether the
discrete tasks and operations. The design process is divided into stages
machine’s output is aligned with our own. These applications are
and packages of information through drawings, models or simulations.
inevitably creeping into the routine of practices, replacing straightforward
The modern digital landscape, through the adoption of computer-aided
activities. They offer many benefits, providing we remain circumspect
design (CAD) and building information modelling (BIM), has integrated
about their limitations. They can be easily sold to an overstretched
the quantifiable whenever possible. Being an architect now certainly
profession as a way to spend ‘more time on what they enjoy’, as was
appears much more systematic and computational, but does that
promised last century with the introduction of CAD.1
make architects suitable for replacement by AI?

1
‘Towards computer-aided building design’, RIBA Journal, 1968.

Tom Holberton
AI Researcher and Lecturer,
Bartlett School of Architecture, UCL

30
RIBA AI Report 2024

Parole Composte, by Natalia Michalowksa, UG21 Bartlett School of Architecture: Co-authoring designs with AI Large Language Models. Image courtesy of Natalia Michalowksa.

Extending generative AI to design-specific tasks is more challenging. As we try to replace larger and larger chunks of the design process
New AI software manufacturers are keen to present intelligent tools with machine-learning we encounter difficulties with its blackboxed
that correspond with a systemised definition of architecture, stage by nature. We have become accustomed to digital technologies bringing
stage, output by output. It might be suggested that these tools can greater legibility to process. Software should bring reliable functions
replace pieces of design thinking. The best arrangement of volumes on with clear inputs and outputs. More crucially, when a building must
a site can be treated as a parametric problem to ‘solve’, where multiple reconcile both quantified and unquantified factors, the designer needs
factors of efficiency, daylight and cost can all be combined. Yet the to be able to interrogate how different calculations are manifesting in
datasets used to build this intelligence are permutations generated by an outcome. Architects cannot locate a building design entirely within
algorithm. This approach can render options upon options, combining one simulation or system and ignore everything else.
and blending multiple simulations into a chain of quantified factors.
Extending this into plan generation can be used to draw and multiply Yet AI and deep learning is a paradigm shift in how we relate to
highly standardised layouts automatically, but with little explanation. computation. It offers none of the algorithmic feedback we have
become accustomed to. It is not written in code as operations or
In design, the optimisation of one factor over another, or where a processes, but merely learns an emergent behaviour from examples.
solution must strike a balance, still requires an architect to understand We have to take on trust that its function will be the same next
each factor’s relationship to the overall design. In the current world week as it was last week and that it has not been altered by a new
of computation, we can depend on the clarity of simulations and context. There is no hard-coded function that explains its behaviour,
mathematical functions to explain both how and why changes occur just probability.
with parametric precision. Machine learning may reveal more solutions,
but the danger of this new kind of combinatorial intelligence is that it The expectations that AI is here to automate, to literally self-act,
obscures how a solution has been generated. As a design develops might suggest this technology can fit neatly into the computational
and becomes subject to inevitable changes from other quantified world we have built over the past 50 years. Design has been
and unquantified factors, an architect and the design team are then transformed by and for software, with many discrete and quantifiable
detached from how these changes impact previous assumptions. tasks that might be replaced. And yet, deep learning is a different
kind of automation, acting in the world but without any function,
If AI is to automate larger portions of the design process, then without a parametric legibility. It does not easily replace or extend
a model will need to be trained on a large dataset of drawn problems the established computational tools without undermining our ability
to construction outcomes. Integrating methods of detailed design to to make complementary good creative judgments.
built solutions could offer great benefits in performance, considering
efficiency of fabrication, waste management and sustainability with a
real evidence base. While there may be companies that can assemble
the necessary dataset of training examples, the model’s methods would
largely remain a black box, unavailable for scrutiny. An AI model will
simply present a fully formed but unexplained solution. Any company
would need to absorb all of the risk of performance and failure, with
plenty of confidence in the probability that it works, and insurance in
case it does not. Without a huge liberalisation of standards, the current
network of competencies, responsibility and checks would have to be
applied retrospectively to a solution, with humans checking the
machine’s work without access to any underlying strategy.

31
RIBA AI Report 2024

The unquantified
The unquantifiable has helped architecture to learn the limitations The crossmodal text-to-image generators such as Midjourney,
of generative processes. In education and practice, we frequently Stable Diffusion and DALL.E provide new large-scale methods of
interrogate where a designer’s agency is located alongside the rendering and visualising ideas, questionably web-scraped from all
computer. This debate is integral to the profession. We are ahead of kinds of creative labour. However, these fantastical images, detached
many other subjects in how to assess processes that combine the from any underlying 3D model, have limited usefulness for spatial
human and the generative. This difference has been stark in the past development. The more unexpected impact of this technology is the
year, when many of the arts and humanities have first been confronted prominence it gives to language as a means to access and manipulate
by ChatGPT, raising profound philosophical questions around imagery through entering prompts. The image caption becomes as
automatic writing and integrity. important as the image itself. Language becomes a new, direct way
not just to describe but to directly instruct image construction.
Much of the work of architects that is most cherished lies in these
gaps between discrete pieces of information. It sits within eternally The technology companies who create the text-to-image models
difficult terms and the particularity of design. It depends on complex see creativity as a ‘zero-shot’ or one-click process open to all. It brings
feedback and foresight: how might a sketch be resolved to a built a quality of synthetic representation to anyone; one which was only
detail or absorb other requirements? There are intuitive pieces of possible previously with extensive visualisation software and skills.
experience and judgment that allow the design to progress and This deskilling might inevitably bring architectural representation
emerge – crucially through anticipation but not determination back into the routine of practices, where the combinations of 3D
of an outcome. It remains unquantified and indeterminate, trading models, description and reference images can quickly communicate
in probabilities. architectural ideas. New AI plugins to 3D-modelling software are
already heading in this direction.
So, if this was unsuited to previous modes of computation, is it
now beyond AI? While generative models, such as text-to-image If we see generative AI as an opportunity of perception rather
generators, still use data, they explore what might have been than cognition, we can train tools that augment creativity by relating
considered previously unquantifiable qualities. Through assessing different data types or moving between exploratory forms of
the probability of different combinations of image content and representation. These can be formal: sketches, diagrams and gestures
captions they build subjective but powerful readings. The training to 3D volumes. They might equally be far more speculative and
method measures the patterns of human deployment of meaning connect other sensory inputs, such as music or language. They can
out in the world to build an understanding of syntax and, possibly, be focused on the specifics of a site, or a typology, or a process.
semantics, behaving more like a form of perception than cognition. AI models can act as creative tools for individual projects that work
Learning is created directly through recognising how examples iteratively rather than offering a single-click outcome.
reinforce or diverge from one another.
To harness this will require more direct engagement in the collection
Methods such as these are crossmodal; they learn the correspondence of datasets and training of models by architects themselves. We now
between different kinds of data and contexts. In this they can capture operate in a data-rich environment and have the means to curate and
the patterns and structures of relationships as are tested and applied generate our own data to direct bespoke machine-learning models.
in real life. They mimic the crossmodality of architecture itself, which These can be part of a design process and are already being
depends on multiple sources of representation to develop and refine enthusiastically tested and scrutinised by new generations of
one idea: through drawings, models, plans and sections. The architect students engaging with AI at a code level.
deploys and develops many parallel methods of representation
to interrogate and refine a single design. AI can learn a ‘centralising tendency’, converging to the average
in any dataset, ignoring the outliers and any exceptional data.
This is inevitably a concern with large-scale commercial tools,
where their mass utility is at odds with the specialist and applied
knowledge of a profession. Writing the right kind of technical email
to a contractor or positioning material junctions correctly in an image
might be impossible for a user to prompt out of a general purpose
model. Identifying and safeguarding specialist data might become
a key objective of professional bodies. Bespoke AI models for case
law or medicine are already being pursued and, with methods such
as transfer learning, more generally intelligent models can be
focused towards a specific domain of knowledge.

The technology companies who create

the text to image models see creativity
as a ‘zero-shot’ or one click process open
to all. It brings a quality of synthetic
representation to anyone; one which was
only possible previously with extensive
visualisation software and skills.
32
RIBA AI Report 2024

A new relationship
From a mixture of quantifiable and unquantifiable factors, design If architecture can engage with this technology as a form of
has become a mixture of the computable and incomputable. There machine-perception, one that perceives a site, a design and a process
are dangers in assuming this relationship with computation remains in novel and unexpected ways, architecture will have a new tool for
the same with AI. If machine learning is adopted only as a technology the unquantifiable. It can extend and recalibrate the relationships
of the quantified, it risks a further systemisation of architecture by and associations that are already out in cities and everyday life
those who operate within conventional metrics. It will not, however, and serve as a constant source of reference for new buildings.
extend transparency or legibility, but risks extending power to software It can combine different patterns of participation and representation,
with inscrutable logic. It will make it harder to reconcile the complex extending the crossmodality of practice to the community. Architects
mixture of factors that combine in a building, frustrating judgment must engage with AI, not only as a means to solve universal problems,
and creativity. but also to localise problems with bespoke models and data. This
should ultimately place architecture at the centre of the debate
about how to use the technology, rather than just being another
automated application.

Cross-modal compositions, by Rolandas Markevicius, PG21 Bartlett School of Architecture: Designing architecture and music simultaneously with AI. Image courtesy of Rolandas Markevicius.

33
RIBA AI Report 2024

Royal Institute of British Architects

66 Portland Place, London, W1B 1AD

+44(0)20 7307 5355

[email protected]
www.architecture.com

34