Floowed vs Zest AI: Credit Decisioning vs ML Scoring

For lenders who want the credit decision owned by their own policy rather than one vendor's model, Floowed is the stronger choice than Zest AI: it runs your credit policy on a deterministic policy engine, weighs any score as one input, Zest's included, and works in any market.

Zest AI is a US-only machine learning credit-scoring engine that has extended into auto-decisioning, fraud, and Model Risk Management documentation. The score is the anchor. The customer base is US banks, credit unions, and auto lenders, and the compliance package (FCRA, SR 11-7) is US-specific. Floowed runs the whole credit case, with no proprietary score of its own. The deterministic policy engine runs your credit policy on every application, with the rules and the source behind each call, and orchestrates any score (FICO, Zest, CredoLab, Trusting Social, in-house) into a final decision. It is fed by your own systems, the bureaus, and documents at any quality, and it chases what nobody has written down yet. Different geography, different center of gravity, different buyer. If you are a lender outside the US, Zest is not the conversation and Floowed is built for you.

Who each is built for

Zest AI earns its place with US banks, credit unions, and auto lenders that want a custom ML underwriting model built on FCRA-compliant bureau data, with the regulatory documentation US examiners expect. The named customers (Citibank, Truist, Freddie Mac, dozens of credit unions) are all in that bracket, and a Zest model can still run as one input to a Floowed policy.

Pick Floowed if you want a decision layer that orchestrates any score rather than locking you into one model vendor. Floowed is global, HQ Singapore, and score-agnostic by design: bring any score, we orchestrate, we don't compete. It takes the application however it arrives, from your systems, the bureaus and documents, and credit and risk teams operate the policy. For the underlying category split, see bringing your own credit model, and what a credit decisioning platform is for the category primer.

Quick comparison

Are they actually competitors?

Only partly, and only in one geography. Zest AI began life in 2009 as ZestFinance, a machine learning credit-scoring company. Sixteen years later the score is still the gravitational center of the product, even as the company has built out auto-decisioning, fraud detection (Zest Protect, August 2024), Model Risk Management documentation, and a Temenos-native embed (April 2025). Zest now markets itself as "AI-Automated Credit Underwriting" and claims to auto-decision 80 percent of applications. That is a real decisioning footprint inside the US.

Floowed went the other way. We built the decision layer first and never shipped a score. The policy engine is where credit and risk teams compose policy: read a payslip, process a bureau report, call a fraud signal, branch on debt-to-income, escalate edge cases for human review. Any score plugs in as an input, including Zest's. We are the orchestration layer, not the model factory.

Geography settles the rest. Zest AI is US-only. FCRA and Model Risk Management compliance, the bureau data they train on, and the credit union channel they sell into are all US-specific. Across most of the world, Zest is not on the shortlist. Floowed is. So inside the US, we overlap on the decisioning story for buyers who want score-agnostic policy logic. Outside the US, we are not in the same room.

What Zest AI does well for US lenders, and where Floowed goes further

For a US bank, credit union, or specialty lender, Zest AI is a serious ML scoring vendor, and the reasons are factual: sixteen years of ML model development on FCRA-compliant bureau data; auto-generated Model Risk Management documentation that maps to SR 11-7 and the related US framework, including the validation, fairness testing, and monitoring artifacts examiners expect; named US customers including Citibank, Truist, Freddie Mac, First National Bank of Omaha, First Hawaiian Bank, and a deep credit union book through VyStar, Suncoast, Golden 1, Hawaii USA FCU, and Idaho Central; native Temenos integration since April 2025, plus MeridianLink and Origence partnerships through credit union service organizations and LOS embeds; and Zest Protect, fraud detection bundled with scoring and distributed through MeridianLink since February 2025.

All of that is built around the score. Floowed goes further on the decision itself. The policy is yours, written as explicit rules on a deterministic policy engine and back-tested against your historical book before it runs. The application is gathered from your systems, the bureaus and documents, and what is missing is chased. Any score, a Zest model included, enters as one input the policy weighs. A US lender whose bottleneck is approval rates on thin-file or protected-class applicants can run a Zest model to move that curve and Floowed to make the call, without tying the decision to one model vendor. Outside the US, the comparison is simpler still.

Zest AI outside the United States

This section is short because the answer is short. There is no Zest entity, sales motion, compliance package, or named customer outside the US. The company has not announced a non-US expansion and the product is structurally tied to US bureau data and US Model Risk Management documentation. A lender outside the US evaluating decisioning platforms is not evaluating Zest. They cannot, given how the platform is built.

That matters for the buyer this comparison is for. The single-largest competitive question for Floowed in the US is whether a Zest customer would be better served by a score-agnostic decision layer on top. The single-largest competitive question for Floowed outside the US is whether a lender should evaluate Zest at all, and the honest answer is no.

Where the Zest model breaks for the buyer outside the US (and for some inside it)

Zest's model-led architecture is the right shape for a US lender whose primary bottleneck is the score itself. For a different buyer profile, four things break.

The score is the anchor, but the score is not the bottleneck for most lenders. Zest's pitch is that a custom ML model on FCRA bureau data lifts approval rates by 25 percent without added risk, and by 30 percent on average across protected classes. Those are real numbers for a US lender whose denial pool sits behind a generic FICO cut-off. They are not relevant numbers for a lender whose denial pool sits behind a credit team manually retyping numbers from a JPEG of a payslip, or for a lender whose policy needs to change quarterly based on portfolio performance and macro conditions, or for a lender who already plugs in CredoLab, Trusting Social, or an internal model and wants the decision layer to orchestrate them. The score is one input. For most lenders, fixing the score is not the bottleneck. Fixing the policy and getting the case complete is.

Document intake is not a primary capability. Zest is built around structured bureau and application data. The platform assumes the inputs are clean and complete: API-fed bureau scores, structured KYC payloads, digital application forms. For a US credit union onboarding through Temenos, that assumption holds. For a lender whose applicants send photos of payslips, scans of bank statements, and handwritten income declarations, with pieces still missing, that assumption is the operational bottleneck. Zest does not solve it. Floowed treats it as a first-class part of the platform: documents are one of three input lanes, and what is missing gets chased.

The model-led operating profile does not fit most credit and risk teams. Zest's policy editing surface is model-led, and built for credit teams that think in models, cut-offs, and challenger experiments. That maps well to a US bank's risk-engineering team. It does not map to the credit and risk teams at a fintech, NBFC, or multifinance lender who edit the policy by writing a rule the way they would explain it on a whiteboard. Writing the rule directly is closer to how the policy is actually written in those lenders, and that is the operator surface Floowed is built for.

One vendor for scoring, decisioning, and fraud is the wrong bundle for a buyer who already has those covered. The Zest pitch packages scoring, auto-decisioning, fraud detection, and MRM documentation under one roof. For a US credit union that wants one vendor across all of it, that bundle is the value. For a lender that already has a bureau relationship, an existing fraud vendor, and a separate KYC stack, paying for the bundle to get the decision layer is the wrong economics. Floowed is one platform that runs the credit case and orchestrates whatever you already have.

Where Floowed is built to win

Floowed was built around four structural choices, each anchored to a buyer Zest does not sell to today.

Score-agnostic by design. Floowed is the orchestration layer, not the model factory. Bring any score. We orchestrate, we don't compete. CredoLab, Trusting Social, Zest, FICO, Experian, your local bureau, your internal model, or a combination: it is absorbed unchanged. Use whichever score is right for your portfolio in your market. The policy engine combines whatever score you bring with bureau data, bank statement analysis, internal application history, KYC results, and the documentary evidence the applicant provides. The score is one input to the policy. We do not compete with scoring vendors and we do not compete with bureaus. We use them.

For a lender without a single dominant national score, this is the only architecture that works. Many markets do not have a single dominant score the way the US has FICO. CredoLab covers some of the alternative-data layer, Trusting Social covers another, the local bureaus cover the traditional layer, internal models cover the rest. A platform that sells you its own score and then bundles decisioning around it is asking you to pick a horse before you have run any of them. A platform that orchestrates any score lets you pick whichever one wins on your portfolio.

It takes the case however it arrives. Floowed does not wait for clean inputs. Underneath is an AI-native, multi-agent runtime, model-agnostic, routing between open-weight and frontier models on cost, latency and data residency. It pulls what already exists from your systems over API and MCP, and takes in external sources such as bureau reports, registry checks and KYC results. It reads documents as the third lane: handwritten passbooks and payslips, photographed and skewed bank statements, scanned business registrations, application forms with handwritten corrections, at any quality. That means income normalization, cash-flow and bank-statement analysis (ADB, DSCR), tampering signals, and figures cross-checked across documents, not just characters off a page. On secured and auto loans, the vehicle title is checked against the chassis and plate photos. What is still missing, it chases from the person, system or agent who holds it, as a long-running asynchronous case that resumes when the answer lands. Identity verification runs through the KYC provider you connect. The output is a complete case the policy acts on directly. This is built in, not a partner integration and not a side capability. If an incomplete case is the operational bottleneck in your portfolio, this is the structural difference that separates Floowed from a model factory built for clean digital inputs. For the deeper mechanics, see bank statement analysis software.

Credit and risk teams are the operators. The policy engine is deterministic designed to be operated by the people who actually own the credit policy. Credit and risk teams run it day to day and own policy authoring. Rules are written as explicit conditions. Each rule holds one condition and the policy combines them: salary below $1,500, requested amount above $5,000, employed less than six months, and when all three hold, the application goes to manual review. The team ships that change directly, after replaying it against the historical book if they choose. Versioning, rollback, and per-decision audit trail are automatic, audit-grade, with the rules behind each call. They also choose how much it decides on its own: every case, only the routine ones, or none. There is no Floowed services team translating the policy into the platform's internal language because there is no internal language to translate into.

Published pricing, live in minutes. Floowed charges for credits: the work done on a case, not seats and not models, at a cent a credit at every size. You can start free with $80 of credits, no card and no sales call, and see your first policy run on your own documents right away. Paid plans start from $100 a month. Custom lending set-ups, with your bureau data, scorecard and integrations, are quoted; the rest is on the pricing page. It goes live in minutes on the platform, with set-up support for lending integrations. That lands well under the large enterprise platforms and their multi-month procurement.

In production at Alon Capital, founder Rene de Jesus puts it plainly: "Floowed reads the documents, runs our credit policy, and surfaces a decision in minutes."

How the case arrives is the structural difference

At the policy layer, credit decisioning platforms can look alike. Both run rules. Both have audit trails. Both can ingest a credit score and turn it into a yes, refer, or no. The difference shows up before the decision logic ever runs: in whether the case is complete when it gets there.

If your applicants send clean digital data through APIs and structured forms, the intake question is uninteresting. The data is structured. You parse it, you decide. US credit unions running on Temenos or MeridianLink usually live in that world. Zest is built for that input pipeline.

Most real-world lending is not that world. The applications include scanned identity documents, handwritten income statements, photographed utility bills, partially completed application forms with handwritten corrections, business registrations from a dozen formats, and bank statements that range from a clean digital export to a phone photo of a printed statement with the corner folded over. That is the input pipeline, and it usually arrives with something still missing: a second payslip, a co-borrower's statement, an answer only the applicant has.

Floowed takes all of that as a first-class product surface. It pulls from your systems, processes the bureau reports, reads the documents at any quality, normalizes income, runs cash-flow analysis (ADB, DSCR), flags tampering, and chases the missing pieces until the case is ready to decide. Then the policy engine runs your policy. No separate document vendor procurement, no integration work, no second contract. If a complete case is your operational bottleneck, this is not a feature you should expect a model factory to retrofit. It is a structural choice Floowed made on day one.

Where each one fits in the stack

If you are a US bank, credit union, or specialty lender whose bottleneck is approval rates on thin-file or protected-class applicants, you have the regulatory appetite and budget for an enterprise SaaS contract, your LOS is Temenos, MeridianLink, or Origence, and you want one vendor for scoring, decisioning, and fraud, Zest AI is on your shortlist for the model, and its score can still run inside a Floowed policy.

If you want a decision layer that does not lock you into one model vendor, the rest of the platform fit lines up with Floowed. Score-agnostic decisioning that orchestrates whatever score is right for your portfolio. Cases taken however they arrive, from your systems, the bureaus and documents, with what is missing chased. Credit and risk teams who edit the policy directly without filing a ticket every time a cut-off needs to move. Live in minutes on the platform, with set-up support for lending integrations, at a published, consumption-based price the budget can absorb. That is what Floowed is built for. Different category, different buyer, different geography from Zest.

What does Zest AI actually cost?

Zest AI doesn't publish pricing, which is normal for custom model development. Contracts are quoted per institution, and industry write-ups and partner material describe typical engagements for mid-sized credit unions and banks as six-figure annual, multi-year SaaS and services agreements that bundle model development, validation support, and ongoing monitoring. Per-decision rates are reported to be negotiated on volume, and scope grows with the loan products covered.

That is not a criticism. A custom-built, validated underwriting model is a serious piece of work, and the price reflects it. The question for a lender is simply whether you are buying that, or buying decisioning infrastructure.

Floowed sits on the other side of that question: it runs the credit case, priced on credits for the work done on a case, with $80 of free credits to start, paid plans from $100 a month and custom lending set-ups quoted. See pricing. A Zest score can ride straight into a Floowed policy as one input among several. We orchestrate scores, we don't build them, so the two line items don't overlap.

Frequently asked questions

Is Zest AI available outside the United States?

No. Zest AI operates in the United States only. The compliance package (FCRA, Model Risk Management) and the bureau data partnerships are US-specific. There is no presence outside the US as of this writing. A lender outside the US is not a Zest customer.

Is Zest AI a scoring platform or a decisioning platform?

Both, with scoring as the anchor. Zest started as ZestFinance in 2009 building ML credit-scoring models. They have since extended into auto-decisioning, workflow automation, MRM documentation, and fraud detection. The product still leads with the model and bundles decisioning around it.

Can Floowed integrate with Zest AI scores as inputs?

Yes. Floowed is score-agnostic by design: bring any score, we orchestrate, we don't compete. A Zest score is passed in over the API with the rest of the application, from your LOS or wherever the score lands, and becomes one input alongside bureau data, KYC results and figures read from documents. Your credit and risk team sets the weight and threshold it carries, back-tests that change against the historical book before it goes live, and every decision records the inputs it used.

How does Zest AI's auto-decisioning differ from Floowed's policy engine?

Zest's auto-decisioning is model-led. The ML model produces a score and the workflow optimizes cut-offs around it, with credit teams tuning policy through that model surface. Floowed's policy engine is policy-led and deterministic. Credit and risk teams compose rules as explicit conditions, branch on any attribute including external scores, and edit policy directly without retraining a model.

What is Model Risk Management and does Floowed support it?

Model Risk Management is the US regulatory framework (SR 11-7 and related guidance) requiring banks to validate, monitor, and document the models they use in lending decisions. Zest auto-generates this documentation for their ML models. Floowed does not produce a proprietary score, so there is no scoring model of ours to validate: the credit call is made by rules your team wrote, on a deterministic policy engine. Every decision records the policy version, the inputs and the evidence behind each check, and a policy change can be back-tested against the historical book before it goes live, which is the governance record rules-based decisioning needs.

Does Zest AI have native document intelligence?

Not as a primary capability. Zest is built around structured bureau and application data. Document intake on handwritten bank statements, scanned payslips, or photographed IDs is not the lane Zest sells into. Floowed treats documents as one of three input lanes, reading and analyzing them at any quality, and chases the ones that never arrived.

How much does Zest AI cost?

Zest AI quotes custom, per institution. Industry write-ups and partner material describe typical engagements as six-figure annual, multi-year contracts bundling model development, validation, and monitoring, with per-decision rates negotiated on volume; confirm current numbers with Zest directly. Floowed prices separately and publicly: credits for the work done on a case, $80 of free credits to start, paid plans from $100 a month, and custom lending set-ups quoted. It orchestrates a Zest score as an input rather than replacing it.

Does Floowed use AI models to make the credit decision?

Not to make the call itself. Floowed runs on an AI-native, multi-agent runtime that is model-agnostic: agents gather the case, read documents and analyze cash flow, routing each task between open-weight and frontier models on cost, latency and data residency. The decision is executed by the deterministic policy engine running your credit policy as written, versioned, with a full audit trail. A Zest score, or any other, enters as one input the policy weighs. The rules that decide are the rules you wrote, and every check carries the evidence behind it.

What about Floowed vs other decision engines?

See Floowed vs Taktile for the modern decision-engine comparison, Floowed vs Provenir for the incumbent comparison, and Floowed vs Scienaptic for another AI-for-credit vendor. The pattern holds across them: where applications arrive incomplete and you want the policy in your own team's hands, Floowed is the stronger fit.

The bottom line

Credit scoring tells you the risk of a borrower. Credit decisioning tells you what to do about it. Zest AI is a strong US pure-play on the scoring side, with auto-decisioning built around its model and a deep credit union and US bank book; the decision itself is where Floowed goes further.

Floowed solves a different problem for a different buyer. Score-agnostic decisioning that orchestrates whatever score is right for your market. Cases taken however they arrive, from your systems, the bureaus and documents, with what is missing chased. Credit and risk teams who edit the policy directly. Published, consumption-based pricing, and live in minutes on the platform, with set-up support for lending integrations. If you are a US lender who wants decisioning that is not anchored to one model vendor, Floowed is the alternative the bundled-vendor pitch does not offer. If you are a lender anywhere else, Zest is not the conversation and Floowed is.

See it on your own applications

If you are evaluating credit decisioning platforms, the fastest way to decide is to see it on your own applications with your own documents. We will show you the policy engine, a live policy edit, and document intake on real applications. Start free, or talk to us.