Embedded lending is credit offered inside a non-financial product, at the moment the customer needs it. A marketplace advances working capital to a seller against their sales history. A logistics platform funds a fleet operator’s fuel costs. An accounting product offers an invoice advance from inside the invoice screen. The customer never visits a bank, and in most cases never sees one named.
The commercial appeal is obvious, and the reason it is harder than it looks is equally simple. Lending is a balance sheet activity with regulatory obligations attached, and building the interface is perhaps a fifth of the work. The other four fifths are capital, credit decisioning, servicing and compliance.
Key Takeaways
- Three questions decide the model: whose balance sheet, whose licence, and who owns the credit decision. Everything else follows from those.
- Platform data is the actual advantage. Transaction history a lender cannot see is what makes the underwriting better than a bank’s.
- Servicing and collections are the part most platforms underestimate, and they are where the economics are won or lost.
- Build the decisioning layer so policy can change without a release. Credit policy will change more often than the product.
- In the UAE, the licensing position must be settled before build, because it determines what your platform is allowed to do and say.
Three questions that define the model
Whose balance sheet funds the loan. Your own capital, a partner bank’s, a debt facility, or a marketplace of institutional lenders. This determines your capital requirements, your risk exposure and your unit economics more than any other decision.
Whose licence covers the activity. Lending is a regulated activity, and the answer is either that you hold a licence or that you operate under a partner’s. Operating under a partner’s licence constrains what you can control, particularly around credit policy and customer communication.
Who owns the credit decision. This can sit with you, with the funder, or be shared through a policy you set and a funder who retains veto. Where it sits determines whether your platform data advantage is actually usable, because a funder applying its own generic scorecard will not use the signals that make your book different.
Embedded lending: credit originated inside a non-financial platform’s own product experience, where the platform owns the customer relationship and a licensed lender, sometimes the platform itself, holds the regulatory and funding position.
Your data is the product
A bank assessing a small business sees filed accounts, a bank statement and a credit bureau file. A marketplace assessing the same business sees two years of transaction-level sales, refund rates, dispute history, seasonality, customer concentration and whether volume is growing or declining month over month.
That difference is the entire commercial case for embedded lending. It supports approving businesses a traditional file would decline, and pricing more accurately across the book. It also supports better collection outcomes, because repayment can be taken as a share of platform revenue rather than as a fixed monthly instalment.
Realising it requires the decisioning layer to actually consume those signals, which means feature engineering on your own data rather than passing a few summary fields to a partner’s scorecard. It also means being able to explain the decision, both because regulators expect it and because customers ask. Our note on designing credit scoring you can actually explain covers the modelling side, and the credit scoring engine page covers how we build it.
The parts that get underestimated
Servicing is the largest one. Once money is out, someone has to handle repayment scheduling, partial payments, early settlement, restructuring, hardship cases, statements and the accounting behind all of it. Platforms that treat servicing as a phase two problem end up doing it in spreadsheets at exactly the point volume makes that impossible.
Collections is the second. Every book has arrears. The question is whether you have a defined process, the ability to restructure, and a legal route when needed. In an embedded context there is an additional decision that traditional lenders never face, which is what happens to the customer’s access to your core product when they fall behind. Getting that wrong damages the platform business to protect the lending business.
Reconciliation is the third and the least visible. Money moves between the funder, the platform, the borrower and the payment provider, and every leg needs to reconcile daily. This is ordinary engineering, but it is unglamorous and consistently under-resourced, and it is the first thing that breaks at volume.
Build decisioning so policy can move
Credit policy changes. It changes when performance data arrives, when the funder revises appetite, when a segment deteriorates, and when the macro picture shifts. If policy is embedded in application code, every change is a release, and the lag between seeing a problem and responding to it becomes the risk.
Separate policy from plumbing. Rules and thresholds belong in a configurable decisioning layer that a credit team can change under governance, with versioning, an audit trail and the ability to run a new policy in shadow against live applications before it takes effect. Every decision should be reproducible, meaning you can reconstruct exactly which policy version and which data produced a given outcome, which matters as much for internal disputes as for regulators.
The UAE position
Lending and credit activity in the UAE sits under the Central Bank, with the DIFC and ADGM operating separate regimes under their own regulators. Which one applies depends on where you are established and who you serve, and the answer changes what your platform may do, what it must disclose and what it may call itself.
Two developments are relevant to anyone building here. The Central Bank’s Open Finance Regulation, gazetted in April 2024, obliges licensed banks and insurers to provide data access and transaction initiation to Open Finance Providers, which over time gives platforms access to bank-held data with customer consent rather than only their own. And Al Etihad Payments, established as a Central Bank subsidiary in 2023, operates the Aani instant payments platform launched in October 2023, which changes what is possible on disbursement and repayment timing.
Settle the licensing question with counsel before the build starts. It is the input that determines the architecture, not a compliance review to run before launch.
Frequently asked questions
Do we need a lending licence to offer embedded credit?
Either you hold one or you operate under a licensed partner’s, and which applies depends on your jurisdiction, your customers and how the arrangement is structured. This is a question for regulatory counsel before you build, because the answer determines what your platform is permitted to control and communicate.
How long does it take to launch?
With a funding partner already in place and a defined initial product, a first live cohort is usually six to nine months, with much of that spent on partner integration and compliance rather than on the application. Building your own decisioning and servicing capability rather than using a partner’s extends it, and is generally worth it only if your data advantage is genuinely central to the proposition.
Should we lend off our own balance sheet?
Rarely at the start. Own-book lending ties up capital and concentrates risk before you have performance data to price against. Most platforms begin with a funding partner, build a track record, then reconsider once the book has enough history to be assessed on its own terms.
What is the biggest technical risk?
Reconciliation and money movement, not decisioning. Credit models can be wrong and the effect appears gradually in performance data. Broken reconciliation between funder, platform and payment provider produces immediate, visible errors in customer balances, and those are far harder to recover from.
How does this relate to buy now, pay later?
BNPL is one form of embedded lending, aimed at consumers at checkout with short tenors. The same architecture questions apply, but consumer credit carries heavier disclosure and affordability obligations than business lending, and the regulatory position is more prescriptive. Do not assume a B2B design transfers.
Building credit into your platform?
Kentro builds decisioning, servicing and reconciliation layers for embedded lending across the UAE and GCC, scoped against your funding model and licensing position.
