SiteKing: AI Construction Inspections That Cut Defect Recording Time by 60%

Client: SiteKing

Industry: Construction Tech & Quality Assurance

Platform: Web app that installs on any phone or tablet

Scope: AI defect detection, offline inspections, migration from no-code

Technology: React, Supabase, Claude AI

Result: 60% less time spent recording defects

Key Results

60% less - time spent recording defects on site

3× faster - defect reports ready to send

200 defect types - recognised by AI from the NHBC standards library

100% offline - inspections work without the internet and sync later

90% confidence - results above this level are accepted automatically, the rest go to a manager

The Challenge

SiteKing helps UK housebuilders find build defects, record them and make sure they get fixed. The first version ran on a no-code platform and couldn’t grow further.

Slow paperwork on site. Photos, notes and details took too long to record by hand.

Inconsistent defect records. Different inspectors described the same problem in different ways.

No signal on site. Construction sites often have poor internet, and inspections couldn’t wait for a connection.

Outgrown no-code. The platform needed its own foundation to scale securely.

Many companies, one platform. Each company’s data had to stay private, with different access for different roles.

Chasing subcontractors. There was no simple way for subcontractors to see their defects and prove they were fixed.

What We Did

Taught AI to spot defects. Inspectors take a photo, and AI matches the problem to one of about 200 NHBC defect types. It then suggests the severity and which trade is responsible.

Let inspectors point and speak. Inspectors can circle the problem on a photo or simply dictate notes, and AI turns them into a tidy inspection record.

Kept a human in the loop. Results the AI is at least 90% sure of are accepted automatically. Anything less goes to a manager to confirm or correct, and the AI learns from those corrections on each project.

Made it work offline. The app installs on a phone or tablet and works without the internet. Inspections and photos are saved on the device and sync automatically when the signal returns.

Automated reports. PDF reports for inspections, trades and directors are generated in a click, plus quality dashboards, defect heat maps and a history for every plot.

Gave subcontractors a simple portal. Subcontractors get a link by email, see their defects, upload proof of the fix, and the site manager is notified straight away.

Moved off no-code safely. We rebuilt the platform on its own infrastructure step by step, without losing any features along the way. Users can sign in with a magic link, a password or Google.

Kept every company’s data separate. Each company sees only its own data, and five access levels control who can do what.

The Result

SiteKing went from a no-code tool to a standalone AI inspection platform that works anywhere on site.

• 60% less time recording defects
• Reports ready 3× faster
• Defects identified from photos across about 200 NHBC types
• Inspections never stop, even with no internet connection
• Defects tracked from discovery to fix, with subcontractors in the loop

About the SiteKing Project

SiteKing is an AI-powered construction quality inspection platform for UK housebuilders. Time2Launch migrated it from Base44 to a dedicated Supabase architecture with PostgreSQL, Supabase Auth and Storage, and Deno Edge Functions, hosted on Hetzner. The React frontend works as an offline-first Progressive Web App.

AI defect detection uses the Anthropic Claude API with vision, matched against a controlled library of NHBC defect types, with confidence-based manager review. Voice notes are converted into structured inspection fields. Row-level security separates data between companies, while subcontractor links, PDF reports, defect heat maps and plot histories cover the full remedial workflow.

The project shows our approach to AI consulting and development: practical AI built into everyday field work. Building field software, moving off a no-code platform or adding AI vision to your product? See our web application development and digital transformation services.

Related case studies: Tener.AI — AI Recruiting Platform · ElderCare — Home Care Platform