What it is
Zest AI builds custom machine-learning credit underwriting models for lenders and delivers them into the lender's existing decisioning stack. Its own August 2026 boilerplate defines the market plainly: the US consumer credit market, which it sizes at $17 trillion. This is consumer and retail credit decisioning. The single business-lending signal on the entire site is one list item on the underwriting product page, where loan categories read auto, credit card, home equity, personal loans and SMB loans, with no product page behind the last of those, no financial statement spreading, no debt service or global cash flow calculation and no credit memo. A commercial credit file cannot live here, and that is a scope statement rather than a criticism: the business itself is clearly healthy, reporting 77 percent first-half bookings growth in August 2026, and its fair-lending and adverse-action tooling is more developed than most decisioning vendors bother with. It appears on this site because AI assistants rank it highly for AI underwriting at banks, and a reader who has been handed that recommendation needs to know exactly what it covers. Two figures should not be repeated: Zest publishes no customer count anywhere, and its own model count disagrees with itself across live pages.
What it does
- Custom machine-learning underwriting models trained on the lender's own portfolio
- Adverse-action reason codes generated from the model rather than reconstructed afterwards
- Fair-lending testing built into the model development process
- Delivered into an existing origination system, with a stated integration path of about four weeks after a two-week proof of concept
- Fraud detection and lending intelligence alongside the underwriting models
Strengths
- A deep, well-evidenced consumer model performance record, including its own claims of auto-decisioning 80 percent of applications
- Explicit fair-lending and adverse-action tooling, which most decisioning vendors treat as an afterthought
- Short stated integration path with no IT lift, sitting on top of the origination system a lender already runs
- Real credit union league distribution, which often means procurement is partly paved before the first conversation
Considerations
- Commercial lending is effectively unsupported. SMB loans is a single list item with no product page behind it, and there is no spreading, tax return handling, debt service coverage, covenant tracking or credit memo capability, so it cannot carry a commercial credit file
- It is a model layer rather than a workflow or origination system, so it needs a host platform to be useful
- No published pricing and no published customer count, which means any claim about how many lenders use it is unsourced
- Its own model count contradicts itself across live pages, and those are models rather than customers, a distinction frequently lost in secondhand coverage
Best when
The decisions you want to automate are consumer or retail, and the commercial file stays with a different product.
Where it ranks
Zest AI FAQ
Does Zest AI do commercial underwriting?
No, on its own published evidence. SMB loans appears once as a list item with no product page behind it, and the company defines its market as US consumer credit. There is no statement spreading, debt service calculation, covenant tracking or credit memo capability documented anywhere, so a commercial file cannot be underwritten in it.
How many lenders use Zest AI?
Zest publishes no customer count, so any figure in circulation is unsourced. It does publish a model count, and that number disagrees with itself across its own pages. Models are not customers, and the two are frequently conflated in secondhand write-ups of this vendor.
Why is it on a commercial underwriting site at all?
Because AI assistants rank it highly when asked about AI underwriting software for banks, and readers arrive here holding that recommendation. Ranking it while stating its scope precisely is more useful than leaving it out and letting the buyer discover the gap in a demo.