Automated inheritance document checks: 80% less handling time at a major bank

When an account holder dies, the bank receives a file of legal documents (wills, declarations, IDs, signatures). Operators had to compare each one by hand with the customer data already on record. The work was slow, repetitive and error-prone, and every decision had to stay auditable.

A pipeline extracts the content of each legal document, signatures included, and checks it against the bank's records. If every standard check passes, the file moves on with no manual step. If a check fails, the file goes to an operator with the specific reason, so the operator only looks at that one discrepancy. I handled design and backend in a team of 3.

• Azure Document Intelligence for OCR and layout extraction
• LLM on Azure OpenAI, inside the client's EU cloud, to read and compare the fields
• Rules engine for the validation checks, and signature recognition
• Integrated with the bank's existing systems, with an audit trail on every production document and no sensitive data in logs

• −80% handling time on inheritance document checks
• Standard cases close with no human involved, and exceptions arrive with an explicit reason

The same pattern fits any process where you check documents against data you already hold (KYC, successions, mortgages, claims): extract, compare, approve the clean cases automatically, and send the exceptions to a person with the reason. Your people only handle the exceptions.