28 hours less on document review.

The practice spent more than 60 hours analyzing every large case file: reading thousands of pages and tracking how names, dates, and evidence connected across them. Notes lived in post-its and Word docs scattered across the team, and cross-referencing 3,000-page files with Ctrl+F was no longer working. Generic AI tools could not handle the scale, structure, and detail sensitivity of a case file — ChatGPT could not cross-reference and cite thousands of pages of witness statements, court decisions, and scanned reports in a way the team could trust for legal work.

I built Casemap — case-analysis software shaped around the way the team reviews large criminal case files. The system ingests the full file — PDFs, scanned documents, handwritten notes, photographs — using computer vision and multimodal AI where needed, classifies each document by type, and indexes everything. The junior associate starts from AI-generated summaries and a working map of parties, roles, relationships, dates, events and communications, then asks questions across the file and receives answers with citations that open the original PDF in the side panel. After review, the system drafts a structured working case summary from the associate's highlights and annotations. Legal judgment stays with the lawyers.

Workflow analysis and pilot observations put the saving at roughly 28 hours of case-preparation time per large case, without adding headcount or reducing human review. Long manual passes through the file became a guided review with cited answers and source verification. At the practice's current case volume, that would roughly double their large-case capacity per year.