HL Hunt will build you a custom underwriting model for a product it doesn’t already cover

This publication has previously covered Hunt Score’s breadth — the more than 55 lending, insurance and eligibility products HL Hunt says its underwriting engine already supports — and its white-label, per-decision pricing. Less examined is a feature the company positions as the answer for everything outside that list: a "Custom AI Underwriting" builder that takes a plain-language description of a credit, coverage or eligibility product and returns a generated application form, document and verification requirements, and a full risk model and decision logic, which HL Hunt says is ready to deploy in minutes rather than the months a bespoke underwriting build would typically take.

The model-management layer sits behind that generation step. HL Hunt describes letting clients build custom risk models tuned to their own portfolio and risk appetite, run champion/challenger tests comparing a new model against the incumbent in live production traffic, and have the system retrain automatically on the client’s own outcomes data over time — a self-learning claim the company frames as continuous improvement rather than a static model shipped once and left alone.

The product page’s own demo decision offers a rare look at how the company frames its composite score. A sample applicant is shown with a 784 Hunt Score and an "Approved" outcome, with the score broken into four sub-signals — payment history at 92, income stability at 87, debt-to-income at 78 and cash flow at 71 — alongside a suggested APR and maximum approved amount. HL Hunt does not disclose how these sub-scores are weighted into the final composite, and the demo is illustrative marketing content rather than a real applicant record.

On oversight, HL Hunt advertises real-time portfolio analytics covering approval rates, default predictions and segment performance, model-drift tracking intended to flag when a model needs retraining before defaults climb, and what it calls examiner-ready audit trails and exportable reports for every decision. Paired with the automated adverse-action notices and ECOA/FCRA-aligned bias monitoring covered in this publication’s earlier reporting on the product, this is the company’s pitch for making a still-new AI underwriting vendor palatable to compliance teams at regulated lenders.

Commercially, HL Hunt separates a no-cost entry point from its paid tiers: the company offers free sandbox access, described as requiring no commitment and no credit card, ahead of the previously reported Starter ($495/month plus $2.50 per assignment), Professional ($1,495/month plus $1.75 per assignment) and Enterprise ($4,995/month plus $0.95 per assignment) plans. A sandbox is a standard way for an underwriting or payments vendor to let a prospective client test integration before committing spend, and its existence does not change the fact that live performance, fair-lending outcomes and model accuracy figures cited elsewhere on the page remain HL Hunt’s own claims, not independently audited results.

The custom-model and self-learning claims raise the same question this publication has posed about Hunt Score before: a generated risk model for a product HL Hunt has never underwritten before is, by definition, untested at the moment it is deployed, and a lender adopting it is relying on the company’s AI-generation process rather than a track record specific to that product. Firms considering the custom-build feature should treat the sandbox as a genuine testing period — validating a generated model against their own historical outcomes data before using it for live credit decisions — rather than assuming AI generation substitutes for that validation.