Every dispute letter triaged the moment it arrives

Borrower correspondence arrives as whatever the borrower had to hand: a clean PDF, a scan, a photograph of a letter taken on a phone. Someone has to open each one, work out what is actually being raised, and decide how urgently it needs to move. That reading is slow, it is inconsistent between readers, and the letter that mentions bankruptcy or military service is the one that cannot afford to sit in a queue.

This workflow runs the moment correspondence lands at its inbound address. It reads every attachment — photographed and scanned images included, not just clean PDFs — strips out the servicer's standard letterhead and disclosures so only the borrower's own substance is considered, then classifies what is being raised: legal action, bankruptcy, disability, death, military service, payment disputes, complaints, fraud, and more. Four independent detection methods run over each document rather than one, so a letter has to evade all of them to be missed.

Each document comes back tagged with its codes, a routing queue, and a priority, assembled into a single spreadsheet emailed to the operator and archived for audit. The output is deliberately privacy-trimmed: the borrower's letter text stays out of it, so the worklist can be circulated without circulating the correspondence itself. Every document is accounted for on the sheet, so nothing goes missing between the inbox and the queue.