RIBA’s 2026 Artificial Intelligence Report records a clear shift in architectural practice. Seventy-four per cent of respondents said their practice was using AI on at least some current projects. The same measure was 59% in 2025 and 41% in 2024.
Those figures are significant, but they need their context. RIBA ran the survey between March and April 2026 and received more than 1,100 responses from built-environment professionals. Participants were self-selecting and did not all answer every question. RIBA says the results should be read as a good indication of AI and the profession, rather than a definitive picture of every architectural practice.
The report therefore does not justify the claim that every firm is already using AI. It does show that use has widened considerably among the people who responded, and that the debate is now moving beyond whether the technology is relevant to practice.
What the 74% figure actually means
The survey asked how often a respondent’s practice used AI in current project work. The 74% figure covers practices that used AI at least sometimes. It does not mean that 74% used AI on every project. It also does not measure whether every use was effective, commercially worthwhile or safe.
The report does show a rise in more intensive use. In 2024, 6% of surveyed practices said they used AI on every or most projects. In 2026, that figure was 31%. Larger practices reported higher adoption: more than 90% of practices with 50 or more staff said they used AI on at least some projects. The comparable figure was 84% for practices with 10 to 49 staff, and 63% for practices with one to nine staff.
RIBA also warns that its survey was updated in 2026 to reflect changes in the AI landscape and make it easier to complete. Where questions changed, year-on-year comparability was lost. The adoption figures use comparable questions, but other statistics should not be treated as a seamless annual time series unless the report says they can be compared.
AI use is moving across several practice functions
RIBA’s accompanying commentary identifies a range of areas in which practices are applying AI. These are not limited to generating imagery or writing text. They include parts of project delivery and the running of a business.
• Early-stage design visualisation supports iteration and communication of options with clients.
• Regulatory compliance checking is being used alongside the technical work needed to navigate complex requirements.
• Practice management covers business processes and routine administrative work.
• Project management concerns coordination and the management of detailed timelines.
• Technical design relates to the development and resolution of building briefs.
Chris Williamson, RIBA President, writes that AI offers “real opportunities” to explore design possibilities more quickly and broadly, create high-quality designs more efficiently, and model outcomes more effectively. His wording describes potential benefits, not a case for removing professional judgement from the process.
RIBA’s Head of Research, Adrian Malleson, makes the same distinction in more measured terms.
Malleson says the report offers clues in three areas: “firstly, the original promise of AI is being partly realised; secondly, there are important and unresolved risks with AI use; and thirdly, the profession and the process of building design will change”.
This combination of opportunity and uncertainty is central to the report. It recognises that AI can speed up existing tasks while asking what happens when the nature of those tasks, and the routes into the profession, are also changing.
Reported productivity is not the whole story
Among respondents using AI, 73% reported that it had improved their practice’s productivity. Twenty-five per cent reported no change and 2% reported a deterioration. When respondents assessed the return on investment from AI adoption, 57% reported a positive return, 41% described it as cost-neutral, and 2% reported a negative return.
The denominators matter. The productivity question refers to survey respondents who use AI. The return-on-investment result refers to respondents who assessed ROI. Neither finding is evidence that every architectural practice has experienced the same effect.
RIBA also captures a more cautious view of the future. Ninety-three per cent of respondents thought AI would have at least a moderate effect on architectural practice during the next five years. Forty-five per cent anticipated a major effect and 15% expected a transformative one. At the same time, only 17% agreed that AI made them more optimistic about the future of architecture, while 49% disagreed.
That gap is important. Widespread expectation of change does not equal confidence that the change will be positive. The report identifies concerns about design imitation and homogenisation, intellectual property, data security and confidentiality, professional indemnity insurance, liability, environmental effects, fee pressure and the future of early-career work.
These are risks and concerns identified by respondents and contributors. They are not evidence that every harm has already occurred. Treating them as settled outcomes would overstate what the survey shows. Treating them as irrelevant would miss a central part of the report.
The early-career issue is a professional question
RIBA highlights a particular concern about the development of early-career professionals. It says AI is currently most readily deployed in areas of architecture that rely less on professional judgement or creative insight and more on standard workflows. These are often the same areas in which people begin to acquire basic professional skills.
Sixty-one per cent of respondents agreed that AI would make it more difficult for early-career professionals to gain the skills they need. Fifty-nine per cent agreed that AI adoption would lead to staff reductions in architecture overall. Yet only 30% agreed that adoption would lead to reduced staff levels in their own organisation.
Again, these are survey views about possible future effects. They do not show that AI has already reduced jobs or training opportunities. The difference between the perception of the profession and the perception of one’s own firm is also instructive. Respondents may anticipate substantial sector-level change while seeing a more uncertain or limited effect in their immediate setting.
The issue is not simply whether routine work can be completed more quickly. It is whether practices still give emerging professionals enough exposure to the tasks, judgement and supervision through which expertise develops. If routine work changes, the learning model may need to change with it.
Adoption is outpacing policy formalisation
The report also asks whether organisations have a formal policy for AI use. Nineteen per cent of respondents said they had a documented policy. Seventeen per cent were developing one, and 15% planned to develop one. Thirty-two per cent said they might need a policy but had no current plans to create one, while 17% said they did not need one.
The figures do not prove that a written policy makes AI use safe, successful or ethical. They do show that formalisation has not kept pace evenly with reported adoption. This makes the content of policy as important as the existence of a document.
A useful policy needs more than a generic statement that staff should use AI responsibly. The RIBA report points to practical areas that need consideration: what information may be entered into a system, how output is reviewed, who is accountable for decisions, how intellectual property and confidentiality are handled, and how professional competence is maintained. In a regulated profession, those questions cannot be handed to a software provider or resolved by a single prompt.
Williamson’s conclusion is direct: “These concerns underscore the importance of professionalism: extensive education, lifelong learning, demonstrable competence, an ethical foundation and a deep sense of professional responsibility.” That is the report’s wider argument. AI may alter the work, but it does not remove the need for people who can make accountable professional decisions.
Malleson frames the issue as integration with professional responsibility, rather than a simple contest between people and software.
Takeaway
The RIBA report is not a simple adoption scorecard. It shows increasing AI use among respondents, some reported productivity and ROI benefits, and a substantial set of unresolved professional questions.
For architectural practices, the practical implication is to move beyond informal experimentation without treating a policy document as the final answer.
• Match policy to real workflows. A useful policy should cover data, output review, confidentiality, intellectual property, responsibility and escalation in the contexts where teams use AI.
• Protect capability development. If routine work changes, teams need to consider how early-career staff will develop the technical knowledge and judgement that the profession requires.
• Keep oversight visible. AI may support analysis, visualisation and administration, but accountable design decisions still need professional review.
• Use the survey carefully. The 74% figure is a strong signal from RIBA’s respondents. It is not a licence to generalise beyond the evidence.
The most useful next question is not whether architecture has adopted AI. It is how practices will make adoption consistent with competence, accountability and the long-term development of the profession.
The Project Flux newsletter keeps the built environment’s AI story grounded in evidence, practice and professional judgement. Subscribe to stay with it.
Links and Stuff
All content reflects our personal views and is not intended as professional advice or to represent any organisation.
1