Redesigning document intake for an accounting firm

Context

Employees needed to download, identify, rename, classify and associate each document with the correct client before beginning the accounting work itself. The result was a slow and inconsistent experience for both staff and clients.

Discovery

• Mapped where documents entered the process.
• Separated steps requiring professional judgement from predictable repetitive decisions.
• Identified where incorrect or incomplete documents created delays.
• Clarified how exceptions should be escalated and which system should remain the source of truth.

Product opportunity

How might we reduce repetitive document handling while preserving review, traceability and professional control?

MVP scope

• Centralised document intake, document type classification and defined field extraction.
• Matching documents to client records and detecting missing or low-confidence information.
• Human review for exceptions and audit logging of automated and manual actions.
• Advanced analysis and autonomous accounting decisions were excluded from the first release.

Key product decisions

• Automate predictable work, not professional judgement.
• Make uncertainty visible instead of silently accepting low-confidence results.
• Treat the exception queue as a primary product experience.
• Preserve the existing accounting system as the source of truth.

Journey before and after

Before: client email or upload -> manual download -> manual renaming -> manual classification -> client matching -> missing-information follow-up -> accounting review.

After: centralised intake -> automated classification -> field extraction -> client matching -> confidence check -> human exception queue -> accounting review.

What I learned

Operational automation creates value when it improves the complete human workflow, including exceptions, corrections and accountability. Automating only the ideal path would have moved the bottleneck rather than removing it.