Computer-Vision Document Digitization Tool for a Compliance Consulting Firm

Overview

Built a computer-vision document digitization tool for a compliance consulting firm that regularly receives boxes of paper records from clients undergoing an audit. The tool scans and classifies incoming documents by type, extracts key fields like dates and reference numbers, and organizes everything into a searchable digital case file instead of consultants manually sorting through physical paperwork. This was particularly valuable during time-boxed audit engagements, where the firm's consultants were previously losing days of billable engagement time to manual document sorting before the actual analysis could begin. Extracted data is flagged for consultant review rather than treated as final, keeping a human check on anything the system isn't confident about. A compliance consulting firm's consultants were losing days of billable engagement time manually sorting through boxes of paper records during time-boxed audit engagements before analysis could even begin. We built a computer-vision tool that scans and classifies incoming documents by type, extracts key fields, and organizes everything into a searchable digital case file, flagging low-confidence extractions for consultant review. We trained the classification model on a sample of the firm's own past engagement documents across several document types, then validated extraction accuracy against a full completed audit's paperwork before rolling it out on a live engagement. Document-sorting time at the start of a typical audit dropped from several days to under a day, giving consultants meaningfully more billable time for actual analysis instead of manual paperwork sorting.