Lending

Lending is a document and decision business. Between application and funding sits a pile of paperwork, including statements, tax returns, appraisals, entity documents and verifications, plus a sequence of judgements about whether the numbers add up. Both are expensive, and both compound with volume.

Where we work

Origination. Application capture, borrower and broker portals, stipulation tracking, and the workflow that moves a file from submission to funding. The win is usually in cycle time: every day in underwriting is pull-through risk.

Credit decisioning. Rules engines, scorecards and model integration, with the auditability that fair-lending and adverse-action requirements demand. Automating the clear approvals and clear declines while routing genuine judgement calls to a human is where most of the value is.

Document intake and verification. Bank statements, tax returns, payroll, entity formation documents, and the cross-checks between them. This is the highest-yield place for LLM-based extraction in lending, and the place where naive extraction fails, because the hard part is reconciling documents that disagree.

Servicing and collections. Payment processing, escrow, modifications, delinquency workflow and outbound borrower communication, including the regulated notices that have to be provably correct.

Portfolio analytics. Vintage curves, roll rates, concentration and performance reporting, built on a data model that serves both operations and the reporting you owe to capital providers.

What tends to matter

Explainability. A credit decision has to be defensible to a regulator, an auditor and the applicant. We build decisioning that records not just the outcome but the inputs and rules that produced it.