Foundations
Architecture first
Case data model, auth and document storage designed before screens, with the document-processing service separated out.
JKP CasePilot: AI case management for federal employee defence law.
A specialist law practice drowning in case files and scanned documents. CasePilot gives attorneys a briefing on each case and reads the documents for them.
- In use
- by the practice
- Phase 2
- proposed
- OCR + AI
- document review
The problem
Hundreds of pages per case, and the facts buried in scans.
Federal employee defence cases arrive as large bundles of scanned documents. Attorneys spent hours reading before they could advise. The practice needed case files organised, briefed and searchable, without a large IT project.
What we built
- Case file management for the practice
- AI case briefings generated from the file
- Document analysis and review, including scanned documents
- OCR for scanned legal documents
- Automated tests and post-deploy checks on every build
Technology
What it runs on.
React / TypeScript
Vite
shadcn/ui
Supabase
Vercel
AWS Textract
AWS Bedrock
How AI is used
Google Gemini generates case briefings and analyses documents. AWS Textract handles OCR on scanned legal documents, and Claude via AWS Bedrock runs in the document-processing service. Every build ships with unit tests and post-deploy smoke checks.
How we worked
A working prototype in the practice first, then hardened for daily use.
Block one
Files and briefings
Case files plus AI briefings, reviewed live by the attorneys on real cases.
Next
Phase 2
Deeper document review and workflow, proposed and priced.
Where it is now
In use, Phase 2 proposed.
The practice uses CasePilot day to day. The next phase is scoped and priced.
Talk to us about an AI productHave a document-heavy process that AI could lift?
Start with a Blueprint: 3 to 5 days, and a plain go or no-go call at the end.