AI in Architecture: RIBA Report Insights (and a Reality Check)

The Royal Institute of British Architects (RIBA) has just dropped its 2025 AI Report, and like many in the industry, I was keen to dive in and understand its key findings. I decided to put the report to the test, tasking four different AI models with breaking down its impact on architects: ChatGPT, Anthropic Claude, Perplexity, and Google Gemini.

The results? Interesting, to say the least.

• ChatGPT and Perplexity? Couldn't handle the report's size, throwing back error messages.

• Claude and Gemini successfully processed the information.

• In my opinion, Gemini offered the sharper, more actionable insights.

You can see a summary of their top takeaways in the carousel below.

However, Here's Where I Have a Problem:

While I applaud RIBA for exploring this crucial topic, I have serious concerns about the report's methodology. Only 500 UK architects responded to the survey, out of a total membership of over 50,000! This equates to less than 1% of the total membership.

What this means is:

• It's not statistically representative of the industry as a whole. Statements like "60% of architects are using AI" become highly questionable.

• There may be self-selection bias in that those who responded are likely to have strong opinions or interest in AI (either positive or negative).

• This survey suffers from incomplete responses in that some questions were skipped, either due to irrelevance or lack of knowledge.

• This survey is a qualitative direction, not quantitative certainty that This disclaimer suggests the results provide directional insight into professional sentiment rather than statistically robust conclusions.

This lack of representativeness significantly undermines the report's value. I worry that such numbers may be misleading practices and I would like to see a more responsible use of the statistics to help practices understand where they are in context with others.

So, what should we actually do with the RIBA Report?

Here are my suggestions for how RIBA should move forward;

My Thoughts: Actionable Steps for YOUR Architecture Practice

Beyond the report's limitations, AI is undoubtedly transforming architecture. Here are a few key takeaways that practices can and should act on:

1. Don’t Just Use AI, Understand Where It Fits in Your Workflow: Start by mapping your current workflows, then assess where AI tools can augment, not replace, existing strengths (e.g., feasibility studies or visualisation).

2. Create a Lightweight AI Policy for Your Practice: Start with a simple internal AI charter. This protects your team and builds client trust.

3. Focus on Augmentation, Not Automation: Leverage AI for speed and insight, but keep your creative and contextual judgment at the core. Understand how Generative Design helps early stage design.

4. Start With Use Cases, Not Tools: Too many architects ask, “Which AI should I try?” A better question is, “What problem am I solving?” Pick tools that solve real bottlenecks, not ones that promise vague disruption.

5. Share and Learn - AI Is Too Big to Navigate Alone: Join peer groups, post your experience (good or bad), and attend webinars.

My Key Takeaway:

The adoption of AI for Architects will only be as good as their holistic Technology Strategy. This involves understanding:

• Hardware

• Software

• Infrastructure

• Processes

• Workflows

• Training

• Change Management + more

/

Generative Design - what else?

My commitment to the industry is to create useful resources and I am following this my making my cheat sheet for ALL Generative Design and Early Stage Design tools free! You can download this here: https://addd.io/product/gd-cheatsheet/

For the full report please see below;

Events

Webinar: Flux Kontext for Architecture and Interior Design

Free Live Webinar on 19th July

Join for a hands-on introduction to Flux Kontext Pro, the AI-powered image generation tool reshaping the world of architecture and interior design.Whether you’re an architect, interior designer, or a CGI artist, this webinar will show you how to:

• Generate multiple design views in seconds

• Automate your concept iterations

• Use ChatGPT prompts to guide visual storytelling

• Save 4–6 hours per project phase

• Ditch the endless render revision cycles

What You’ll Learn:

• What is Flux Kontext and how it works

• Crafting visual prompts that align with your design language

• Creating cohesive views (interior & exterior)

• Post-processing tips using AI

• Real-world case studies from our studio

Bonus:

At the end of the webinar, Neman will share exclusive resources and prompt templates.

Catchup on the latest Workflow Webinar 

I held the latest webinar last week:

Smarter Revit Workflows for Architects and BIM Managers!

We explored three game-changing tools that help you design, document, and deliver faster and smarter.

🧠 Pele – Your AI Assistant in Revit
Talk to Revit like you talk to ChatGPT. Pele uses AI to automate repetitive tasks - views, dimensions, overrides, and more. First 20 prompts are free!

🎨 Vyssuals – Real-Time Data Visualisation
Say goodbye to clunky exports. Vyssuals turns your Revit model into a live dashboard -instantly highlighting errors, performance gaps, or clashes through visual feedback.

🤝 D.TO – Collaborative Assembly Design
Design with confidence. D.TO combines AI, BIM, and a mentoring layer to help you manage complex detailing and product selection - all inside a collaborative interface.

💡 Whether you're an architect trying to simplify early-stage design or a BIM manager looking to optimise documentation and QA, this session is for you.

AEC Tech Jobs

Want to Move into ConTech? Check out AEC Tech Jobs - all the best new roles from Computational Designer, Product Manager, Software Developer, Marketing and Sales are all here. Join one of the fastest growing industry out there for ConTech roles.

Are you a ConTech company- reach out and we would be happy to add your latest roles.

Please forward this to a colleague that you think will benefit from the content.

Get in touch with any questions you may have: [email protected]

Thanks for reading!