AI Is Already In Most Architectural Practices. Is It Doing The Right Work?
An Article By: Kristian Brown, Founder, Built Logic
A practical test for finding where AI can improve efficiency, protect margin and give architectural practices better control.
Most architectural practices are already experimenting with AI. The question now is where is it actually worth using. This article sets out a practical way to identify the parts of a practice that are worth changing, where the commercial value sits and how to test one use case properly before committing..
RIBA’s 2026 AI survey found that 74% of practices now use AI in at least some projects. The question is no longer whether architects will use AI. It is whether they are using it in the right parts of the business.
Built Logic’s AI Opportunity Map for the Built Environment
We’ll assess your practice and identify the three areas where AI could have the biggest commercial impact.
Start With The Work, Not The AI
The common mistake is to start with software. Someone sees an impressive demonstration, a team member opens an account or the practice buys another platform before anyone has agreed what needs to improve. That usually creates more experimentation and more to manage, not a better business.
Start with a specific task instead. AI is usually most useful where it can prepare, organise, compare or retrieve information, while the person in charge keeps control of the final decision and submission ready output.
Four Questions To Ask First
Before spending money or changing a process, put the task through four tests.
1. Does it happen often enough?
A task repeated across every project is usually more valuable than a clever solution to something that happens once a quarter. Look at fee proposals, design briefs, meeting records, reports, information requests from contractors and document searches for specs.
2. Is there enough value in improving it?
Saving a few minutes is not enough unless you are doing it a hundred times a week. Look for work that affects scope drift, response time, missed opportunities, rework, capacity or senior leadership team oversight.
3. Is the source information good enough?
The output will only be as useful as the information behind it. Approved templates, previous proposals, project records, standard clauses and decision logs are far more useful than scattered or outdated files.
4. Can someone check it quickly?
A good use of AI should produce a first pass that is quicker to check than to create from scratch. If the result needs rebuilding, or an error could create serious contractual, safety or professional risk, it is gated in a way that it cannot fail.
Where This Can Help In Practice
Fee proposals are an obvious example. A first draft can be built from the client brief, relevant RIBA Plan of Work stages, standard deliverables, exclusions and comparable past appointments.
Meeting and coordination records are another. For the studio, the contractor and any other stakeholders. Notes and approved transcripts can be turned into actions, owners, dates and decisions for everyone to action.
Practices also lose time searching for information they already hold. Previous details, specifications, planning submissions and lessons from completed work can be made easier to retrieve.
The same applies to client reporting. Approved progress information, outstanding decisions and approaching deadlines can be pulled into a first draft, before the project architect issues.
Test One Use Case First
Do not roll a new system across the whole practice because the first test looks promising. Prove it on live work first. Rolling it out too early is one of the biggest mistakes you can make.
Record how the task performs today, test the new process on a small number of examples and compare the result. The diagram below gives a simple structure for doing that.
Set the rules before it starts: which tools are approved, what information can be entered and who checks the output. Expand only when the improvement is clear and repeatable.
