Financial Services · Loan Origination
Auto Finance Lender
Engineering an AI system to accelerate loan origination, surfacing fraud patterns and underwriting misses, for a publicly traded auto lender
Impact
Live in 1 week
1 week
From kickoff to agentic fraud detection
93%
Cost reduction for loan origination
5 mins
Agentic turnaround time
The challenge
Analysts manually reviewed 50k+ applications from dealers monthly, cross-checking W2, 1099s, address proofs, IDs, and bank statements, with 3 human touchpoints for each approval, and 8h manual turnaround time. Income verification from unstructured documents was slow, inconsistent, and missed fraud patterns.
The solution
Distyl built a document intelligence system that ingests applications via API, extracts income data, verifies consistency across sources, and flags fraud or compliance issues. The system learns from analyst decisions, improving accuracy using the lender's own job aids.
AI capabilities
- 01
Document intelligence across sources
Verifying consistency across multiple document types from different sources
- 02
Structured extraction from unstructured data
Computing quantitative fields from unstructured documents using SME-defined logic
- 03
Risk signal detection
Identifying patterns across documents that rules-based and manual processes miss