The problem
An innocence project reviewed hundreds of case inquiries annually to identify potential wrongful convictions. Screening involved reading lengthy case files, extracting key facts, comparing evidence claims, and summarizing findings in evaluation memos. This work was repetitive but required legal judgment.
What we did
- Identified discrete high-priority tasks that drained resources and were feasible targets for AI assistance
- Built an AI tool to extract more than 100 pieces of information from a new client questionnaire packet
- Created AI-generated preliminary screening memos to summarize across volunteer reviews with flagged inconsistencies and potential leads
- Designed attorney review interface to accept/reject AI findings and add judgment-based notes
- Integrated new tools with internal data collection forms to minimize manual data entry
What changed
- Reduced time to review pro bono work product by several hours per case
Built with
Claude API
Clio
agentic workflows
document processing