Document Intelligence for a Professional Practice

The problem

Every engagement meant chasing clients for documents, reading and sorting what arrived, retyping figures into working papers, and answering the same research questions from memory and folders. The firm's knowledge lived in PDFs, inboxes and a few senior heads.

What we did

We automated the document layer: client uploads and email attachments are OCR'd, classified by document type, and their key fields extracted into structured records; missing-document requests and status updates are drafted automatically; and a knowledge assistant grounded in the firm's own precedents, past positions and reference material answers questions with the sources cited, drafting client replies for a professional to review.

What we connected

• Email & client uploads
• OCR & document AI
• The practice management system
• The firm's document store
• Claude
• Review queues in the team's tools

What we automated

• Document intake, OCR and classification
• Field extraction into structured records
• Missing-document chasing and status updates
• Grounded answers with sources cited
• Draft client replies queued for review

Reliability & controls

• Unclear documents routed to a person, not guessed
• Answers cite sources or say "not found"
• Role-based access to sensitive files
• Professional review before anything is sent