Precedent research in small architecture practices is a three-week manual grind because there's no system—only the founder doing it from scratch on every project.
Over 70% of RIBA-chartered firms have fewer than 10 people. In most of these, the principal architect is the bottleneck. They spend 30–40% of project time on non-design tasks: research, documentation, admin. A typical Stage 1 brief for a residential or mixed-use project means reviewing 15–40 precedent projects manually. Dezeen. ArchDaily. RIBA Library. Planning portals. Historic England records. Conservation area maps. All separate sources. All queried by hand.
Then there's the planning system itself. England has 333 Local Planning Authorities. Each has its own portal, local plan, design guides. A practice working across multiple councils cross-references planning history, listed building constraints, and precedent approvals in separate browser tabs for hours. There's no unified tool for this. It's tribal knowledge executed manually each time.
Why generic AI tools don't solve this.
Generic AI tools like ChatGPT or Perplexity can search the web quickly. But they don't know which planning portals matter for your practice's typical project geographies. They don't know your preferred reference sources. They can't apply your aesthetic criteria or client brief template. They surface results that need heavy manual curation before a client sees them. They automate typing. They don't automate judgment.
Architecture-specific AI tools (Maket, TestFit, others) focus on design output and space planning. They assume research has already happened. Practice management software tracks billable hours lost to research but doesn't reduce the research itself. None of these address the intake workflow that eats three weeks.
How this actually gets faster.
The research doesn't take three weeks because architects are slow. It takes three weeks because the process is repeated from zero on every project. A practice's research criteria—preferred LPAs, aesthetic references, material palettes, typological specialisms—live only in the founder's head. When you encode that logic once, the work becomes repeatable and automated forward.
What this means in practice: you send a site address and project brief. An agent queries the relevant planning portals, cross-references conservation area maps, compiles constraints, and pulls precedent references matching your firm's standards. The output formats to your house style and presentation template. You receive a structured constraints brief and precedent pack ready to review and send to the client. Not a starting point. Finished work.
For a three-person practice, this shifts three weeks of unbillable research time into three hours of output handling. The planning data landscape in the UK is becoming more machine-readable through government digitisation initiatives. Early-stage, but directionally significant. A configured agent built around your specific practice logic and typical project geographies becomes a durable operational advantage.
How we handle this at Sprigly.
We configure an agent around your practice's research logic—sources, criteria, output format, project geographies. You brief once. The agent runs the workflow at every project intake. You don't learn a new tool or rebuild your process. Output lands in your inbox formatted as if you'd spent the time doing it yourself, but weeks of work compressed into hours.
This is not generic software. It's a working service built around how you actually work and how you talk to clients. One setup, then monthly access. Each output only needs a final review before it goes out.
If precedent research is consuming three weeks of your time every project cycle, we should talk.
Book a free 20-minute call to walk through what this looks like for your practice. Or email john@sprigly.co.uk.