All work
Document Automation Pipeline
- Client
- Accounting practice
- Sector
- Professional services
- Service
- Automation
- Timeline
- 9 weeks
- Year
- 2024
- Stack
- Python / TypeScript / PostgreSQL / Claude API
Month-end meant a week of copying numbers from PDFs into spreadsheets, every single month.
The problem
Client statements, invoices, and receipts arrived in a dozen formats. Staff normalized them by hand into ledger entries, a process that consumed the better part of a week and produced subtle, expensive errors.
Our approach
We mapped the close process end to end, then automated the document-to-ledger path: classification, extraction, validation against accounting rules, and a review queue for anything that failed a check, with accountants approving, not retyping.
What we built
A pipeline that turns inbound documents into validated, reconciled ledger entries with full audit trails. The team reviews exceptions over coffee instead of transcribing data into the night.
The results
4 daysCut from month-end close
98.6%Extraction accuracy after validation rules
100%Of entries traceable to source documents