The problem
After every inspection a surveyor has a pile of rough notes and photos that has to become a long, tightly structured RICS report. Doing that by hand takes hours, it is repetitive, and it often happens at night. Signal on site is patchy, so a tool that needs a connection is no good. And any AI help has to be traceable, because clients, insurers and RICS can ask how a report was produced.
What we did
We built two apps on one Supabase database. The iOS app (SwiftUI) is for the inspection: dictation, photos filed against each section, and condition ratings, laid out to the RICS Home Survey structure for Level 2 and Level 3. It works offline and syncs when there is signal, and if the phone and the web both changed something, the unsent work on the phone is kept. The Next.js dashboard turns site notes into a draft report in the background, section by section. Surveyors can rewrite any section with AI, upload a past report so the rewrites follow their own writing style, and export a branded PDF. Every AI edit is logged with the section, time and model, and the trail can be downloaded per survey. Around that we built the business side: plans with survey limits and overage, Stripe billing on the web and App Store purchases checked on the server, per-seat firm accounts with a shared survey pool and firm branding, an admin centre with AI cost tracking, and a report builder so new report layouts can be made and versioned without new code.
What changed
- Offline-first iOS capture for RICS Level 2 and Level 3 surveys, with photos and condition ratings filed by section.
- AI drafts made in the background, plus an exportable AI audit trail for every survey.
- Firm accounts with per-seat billing, a shared survey pool and the firm's own branding on PDFs.
- Stripe and App Store subscriptions on one set of plan limits, with purchases verified on the server.
- Over 1,500 commits since January 2026, with k6 load tests run before launch.
