HL Hunt lets lenders describe a credit product and AI builds the underwriting flow
Beyond its Hunt Score decision engine, the AI Underwriting platform includes a generator that produces application forms, verification steps and risk logic for custom credit, insurance or eligibility products.
FinCrunch covered the headline performance figures behind HL Hunt Financial's Hunt Score decision engine in August: sub-250-millisecond decisions, more than 1,000 data signals per application, and the company's stated 34% lift in approvals against bureau-only scoring. A less-examined part of the same AI Underwriting platform at hlhunt.org is a feature the company calls Custom AI Underwriting, aimed at lenders whose credit, insurance or eligibility product does not fit one of the platform's 55-plus pre-built categories.
According to the product page, a lender describes the product it wants to offer and the platform generates the underwriting flow around it: the application form, the documents and verification steps required, and the underlying risk model and decision logic. HL Hunt markets this as ready to deploy "in minutes" rather than the typical build cycle of designing an application flow and risk model by hand.
The platform also allows lenders to build and run their own custom risk models independent of the generator — setting approval thresholds by product type, and running newly built models alongside existing ones in production through what HL Hunt describes as champion/challenger A/B testing, a standard technique in credit risk management for validating a new model against an incumbent on live traffic before fully replacing it.
HL Hunt's own comparison table on the page sets its stated capabilities against both bureau-only scoring and manual underwriting: where it lists bureau models as "instant score only, no decision" and manual review as taking 24 to 72 hours, it lists its own sub-250-millisecond figure as covering the full decision, not just a score. The company also states its models are self-learning, retraining on a lender's own outcomes data over time.
Distribution is built around white-labeling: HL Hunt describes an embeddable decisioning widget that can be dropped into a lender's existing web or mobile app with a single script tag, a full REST API for headless integration, and multi-tenant support for running different models across different products, channels or partners. For a lender evaluating any of this, the standard questions apply regardless of how quickly a flow is generated — what validation HL Hunt can provide on the generated model's fair-lending performance for a specific borrower population, and how a self-learning model's drift is monitored once it is live on a lender's own traffic rather than in the sandbox.