Archibots mini project utilising

This project started as a bit of fun — a chance to explore, stress test, and push boundaries. So far, it includes a set of architecture-focused ChatGPT chatbots and a Drawing Discrepancy Checker - visit archibots.co.uk

Chatbot use is entirely at your own discretion, and users are responsible for any decisions made based on its outputs. All chatbots are powered by OpenAI GPT.

ChatGPT AI assistants

Archibots is a growing set of GPT-based AI assistant chatbots developed using only publicly available documentation. Whilst these tool should not be used blindly (as any other AI tool for that matter), the chatbots offer a glimpse into how AI might begin to support architectural and construction professionals in day-to-day tasks.

ChatGPT - UK Planning

Using a curated knowledge base of 75 planning and legislative documents, this assistant supports queries on national planning guidance and frameworks.

ChatGPT - UK Building Regulations Approved Documents

This AI assistant helps you quickly navigate through over 2,200 pages of regulatory documentation, providing instant answers with specific references to:

• Approved Documents (UK)
• LABC resources
• NHBC guidance

Each response includes direct citations to the source documents, making it easy to verify and follow up on the information provided.

ChatGPT UK - School Design Specification DfE

School design specification has over 300 pages for architects to navigate, remember and incorporate into their designs. And that is on top of all the other regulations the building has to meet..

This Chatbot helps to speed things up by utilising publicly available DfE documentation, and clearly showing where each answer comes from - making it easy to review and validate.

ChatGPT UK - BIM Framework

This chatbot supports general understanding of the UK BIM Framework. Knowledge base used is UK BIM Framework - information management according to BS EN ISO 19650.

*Key terms
- All chatbots are powered by OpenAI GPT. 
- You'll need a GPT account (a free one works fine). 
- We cannot access your chat history.

Why are we doing this?

We’re not saying these AI bots are flawless. But the only way to uncover the true strengths and weaknesses of AI in our field is to try them out in real-world situations.

Truth is, many are hesitant. People don't trust others to use AI responsibly.

The same challenges keep surfacing: Is it really possible to define meaningful, practical use cases? How do we genuinely integrate AI into QA processes when it’s rarely clear where it fits? Ultimately, it’s the designer or firm—not the AI—that remains accountable for the work.

There’s a lot more to discuss. And while image generation grabs most of the headlines, it might not be the area where architects should focus their AI efforts.

That said, we are starting to find real value through testing.

Real-world examples

Planning appeal – "Unwinnable" case

In a tough planning case, where the application for new structures in green belt has been refused and the architect was going for revised application resubmission, new arguments were required to present a stronger case. Since the refusal, the NPPF has been updated, so the architect employed AI to assist with preparing the new application and improving the argument by comparing the withdrawn and new NPPF policies.

The planning chatbot suggested three arguments. Whilst two were already previously used and rejects, the third one opened a compelling line of reasoning. The architect dug deeper, verified the claim, and built a stronger case based on the AI's suggestion.

The AI proved to be valuable in this challenging scenario.

ChatGPT - UK Planning

DfE School design compliance

An architect has been challenged by the project stakeholder on the proposed corridor widths for the new school. The proposed width was non-compliant with DfE spec, which made architect question their design and the change in corridor width would have affected large portion of the design.

However, when architect outlined explained the situation to the AI chatbot, it found out that the project manager was referencing withdrawn guidance, and the proposed corridor width was in fact correct. The Chatbot provided all the references to where the outdated corridor widths were listed and where the new guidance is. This made is simple for the architect to explain their case without the scrolling hundreds of pages of documentation to justify their design decisions.

The result? Faster resolution of design queries (and potential future claims) and great feedback from users.

What's Next?

We're exploring:

• AI workplace assistants for architectural teams
• Multiple topic-specific GPT bots
• Bid preparation assistants trained on company-specific historical data
• And a few other things we can't talk about yet 😉

In conclusion

AI is powerful when used responsibly. It's not about replacing professionals — it's about augmenting our work. We're just getting started.

Have a GPT account? Try it out. Explore. Break things (gently). Give feedback.

Visit archibots.co.uk