Built an AI workflow that chases missing borrower documents, checks completeness, and keeps loan officers updated automatically.
Challenge
A mortgage team was slowed down by document collection and borrower follow-up. Processors manually checked uploads, chased missing pay stubs and bank statements, renamed files, updated loan officers, and re-requested unclear documents. Bottlenecks delayed underwriting-ready packages.
Workflow approach
We built a document collection automation that tracks each borrower's required checklist, sends reminders, classifies uploaded files, validates date ranges and legibility, flags missing items, and updates the loan record. Loan officers see a live status summary instead of asking processors for updates.
Observed outcome
File completion accelerated 42%, processors saved 19 hours per week each, and borrowers received clearer reminders. The team submitted cleaner packages to underwriting with fewer avoidable back-and-forth loops.
Outcome metrics
- Faster file completion
- 42%
- Automated reminders
- 73%
- Saved per processor
- 19h
- Document match accuracy
- 96%
Systems involved
- GPT-4V
- Python
- Loan Origination API
- Twilio
- Google Drive
- PostgreSQL