Rishi Khosla and OakNorth’s technology-enabled credit model

OakNorth’s public model uses software and sector data to support commercial lending while keeping people responsible for structuring, challenge and approval. Co-founder and chief executive Rishi Khosla has positioned the bank around established, growing businesses that can be poorly served by standardised small-business products and too small for the most efficient attention of large corporate lenders.

The useful lesson is not that an algorithm can replace a relationship manager or credit committee. It is that technology can organise evidence, test scenarios and make specialist work reusable. Commercial lending still depends on the borrower’s plan, sector, cash flow, security, management and the lender’s appetite for risk.

The problem was shaped by an earlier borrowing experience

OakNorth says Khosla and co-founder Joel Perlman struggled to obtain suitable finance for their earlier data-analytics business. The current OakNorth company history presents that experience as the reason for building a bank focused on entrepreneurs and lower mid-market companies.

This is a specific segment choice. A growing hotel group, healthcare provider or manufacturer may need a facility shaped around acquisitions, property, seasonality or expansion. A fully standard product can miss that context, while a traditional bespoke process can be slow and expensive to deliver.

For a founder, the relevant product question is whether software can reduce the cost of understanding complexity without pretending the complexity has disappeared. A lender has to know which information can be structured and compared, and which judgement remains particular to the borrower.

The public sources reviewed here do not identify OakNorth’s first credible borrower or explain how that relationship was won. They also do not isolate which customer, hiring, regulatory or financing advantages depended on London. OakNorth describes offices in London and Manchester, while its later US expansion shows that the operating model is not confined to the capital. The first-borrower evidence and a balanced account of London’s contribution remain reporting requirements.

Software makes sector knowledge more reusable

Commercial credit teams gather financial statements, forecasts, management information, security details and market evidence. OakNorth describes a credit-intelligence approach that combines this information with sector and scenario analysis. The system can help a team ask consistent questions and examine how a borrower might behave under different conditions.

That is different from approving a loan through one opaque score. A technology-enabled process can support segmentation, document work, scenario generation and monitoring while a named person remains accountable for the credit decision. It can also preserve prior analysis so a new case does not begin from a blank page.

The OakNorth 2025 annual report describes its model and results from management’s perspective. Its figures include definitions and adjustments that matter. Readers should use the full report rather than extract one performance number as proof that software alone produced the outcome.

Human judgement remains central to structure and challenge

A commercial loan is more than an approved amount. Covenants, repayment, security, drawdown conditions and information rights shape how the risk is managed. Management quality and the credibility of a growth plan also resist simple measurement.

The lender therefore needs people who can challenge forecasts, understand the borrower’s sector and negotiate a structure both sides can operate. A model can identify sensitivities, but someone must decide which scenario is plausible and what evidence would change that view. The borrower also needs a person who can explain a concern and revise the facility rather than receive an unexplained rejection.

Technology is most valuable when it improves that conversation. Faster information retrieval, comparable sector cases and visible assumptions can give a credit officer more time for judgement. Automation that simply increases application volume can shift the bottleneck to review or monitoring.

A bank must connect speed to prudential discipline

OakNorth Bank appears on the Bank of England’s current list of PRA-authorised firms under firm reference number 629564. Its legal information states that it is authorised by the PRA and regulated by the FCA and PRA. That status brings capital, liquidity, governance, risk-management and reporting responsibilities beyond those of a software vendor.

A lender can make a decision quickly only if the evidence and controls remain sufficient. Commercial pressure to originate should be balanced by independent risk challenge, portfolio limits and monitoring. The quality of a model is not demonstrated by approval speed. It is demonstrated over a credit cycle through decision quality, loss experience, customer treatment and the ability to respond when conditions change.

The current PRA supervisory statement on model risk management, published and effective in April 2026, applies to UK-incorporated banks, building societies and PRA-designated investment firms with approval to use internal models to calculate regulatory capital. This draft has not established whether OakNorth is in that scope, so the statement is not evidence that it governs OakNorth’s particular credit tools. It is included only as an authoritative description of model-risk disciplines for firms that are in scope.

Public results need their definitions

OakNorth’s 2025 report identifies Khosla as co-founder and chief executive and reports revenue, profit, lending and credit-performance measures. The report also provides footnotes defining several cumulative or adjusted figures. Those definitions are part of the claim.

For example, cumulative facilities granted are not the same as the current loan book, and estimated jobs or economic value depend on a methodology. Principal loss measures depend on the numerator, denominator and period. A founder profile should not repeat the largest number without explaining what it represents.

The public record does show a bank operating at meaningful scale in the UK and expanding in the United States. It does not reveal every declined application, restructured facility or internal model decision. Independent borrower and credit-risk perspectives are necessary to assess service and judgement.

Expansion tests whether the process travels

OakNorth has expanded its lending activity in the United States. The lower mid-market problem may exist in both countries, but the legal, credit, banking and competitive systems differ. A reusable analytical capability can help, while origination, documentation and portfolio management still need local expertise.

This makes expansion a test of what the company has actually built. If value rests entirely on a particular UK relationship network, it may travel poorly. If sector analysis, workflow and monitoring improve local bankers’ decisions, the capability may be more portable.

The company’s annual report says a material share of 2025 originations came from the United States. That is management-reported evidence of activity, not yet a complete cross-cycle test. Results should be followed over time with the same definitions.

What fintech founders can apply

Choose a narrow customer whose complexity is expensive to understand. Break the work into evidence collection, analysis, judgement, approval, documentation and monitoring. Use software where it improves consistency or makes specialist knowledge reusable. Keep a named owner for the decision and the explanation.

Build the control system with the product. Version models, record assumptions, test outcomes and define when a person overrides a recommendation. Treat adverse cases and changing conditions as essential product data rather than exceptions that spoil the story.

Publish measures with enough context to evaluate them. Speed, originations, revenue and losses answer different questions. A technology-enabled lender deserves confidence only when commercial growth, customer treatment and credit discipline can be considered together.

Limits of the public record

OakNorth’s website and annual report are primary company sources and contain management’s presentation of the model. Regulatory records establish status, not the quality of every loan. Public accounts provide aggregate results but cannot show how much a particular analytical tool changed a decision.

Before publication, this profile needs verified first-borrower evidence, an independent borrower, a commercial credit specialist, evidence about London’s contribution and limits, and a response from OakNorth and Khosla. Current roles and 2025 measures must be checked again. The supported conclusion is that OakNorth describes a combination of technology and human commercial-credit work. A stronger claim about causal performance would require transaction-level evidence that is not public.