/

Underwriting

Underwriting

Loan underwriting is the process of assessing a loan applicant's risk and deciding whether, and on what terms, to lend to them. An underwriter or a decisioning engine weighs income, credit history, and affordability against the lender's policy. Loan underwriting is distinct from insurance underwriting, which assesses a different kind of risk entirely.

Underwriting sits inside loan origination, between document verification and the final decision: it's the risk-assessment stage, not the whole origination journey. Historically it meant a human underwriter's judgment call on a file, but standard applications increasingly go through automated credit decisioning engines instead, with manual underwriting reserved for complex, high-value, or edge-case applications a model can't confidently assess alone.

How does the loan underwriting process work?

Underwriting begins once an application and its supporting documents have been collected and verified. The underwriter, human or automated, pulls the applicant's credit history and score, checks income against declared or verified figures, and calculates affordability: how much of the applicant's income is already committed elsewhere, and what capacity remains for a new repayment.

That assessment is measured against the lender's underwriting policy, which sets acceptable risk bands, loan-to-value or debt-to-income limits, and any conditions that trigger extra scrutiny, such as self-employment income or a thin credit file. The output is a decision, or a set of conditions the applicant needs to clear before one is issued, such as further documentation or a guarantor.

Standard, well-documented applications typically move through this in minutes to hours when handled by an automated decisioning engine. Complex applications, thin credit files, or high-value loans usually still go to a human underwriter, which takes longer but allows judgment a model can't apply on its own.

What do underwriters actually check?

An underwriter checks four things in most loan applications: income, credit history, existing debt, and the specific risk factors tied to the loan type. Income verification confirms the applicant earns what they've declared, usually against payslips, bank statements, or tax returns. Credit history and score show how the applicant has handled debt previously, including any missed payments, defaults, or existing delinquency.

Existing debt feeds into a debt-to-income calculation: how much of the applicant's income is already committed to other repayments, which sets the ceiling on what a new loan can safely add. Loan-specific checks vary. A mortgage underwriter checks the property's value against the loan amount; a business lender checks cash flow and the business's own financials rather than an individual's income alone.

Underwriters also check for anything that doesn't add up: inconsistent information across documents, an income pattern that doesn't match the stated employment, or a debt load that isn't reflected in the credit file, since these are the cases most likely to need manual review even in an otherwise automated process.

Underwriting vs credit decisioning: what's the difference?

Underwriting is the risk assessment: gathering and weighing the evidence about an applicant's ability to repay. Credit decisioning is what happens with that assessment: applying rules or a model to reach an approve, decline, or refer outcome. In a fully manual process, one underwriter often does both in their head. In an automated process, the two are frequently split into separate systems: an underwriting layer that verifies and scores the application, and a decisioning engine that applies the lender's policy to produce the outcome.

The distinction matters because it's usually the decisioning layer, not underwriting itself, that a lender can safely automate first. Automating decisioning on standard, well-understood cases frees human underwriters to spend their time on the complex or high-value applications where judgment still adds value, rather than splitting attention evenly across every application regardless of its difficulty.

Vendors in this space usually specialise in one side or the other: some build decisioning engines, others build underwriting models, and a few build both together. Our lending guide maps which vendors cover underwriting and which cover decisioning across the stack.

How AI changes loan underwriting

AI has moved underwriting from a fully manual, judgment-led process towards one where a model handles standard cases and a human handles the rest. Machine-learning underwriting models can weigh far more variables than a manual policy typically checks, and score applications more consistently across a portfolio, which is why they've become the default for high-volume, low-complexity lending such as personal loans and standard mortgages.

That doesn't remove the human underwriter. Complex applications, thin credit files, and high-value or non-standard loans still go to manual review, both because a model may lack the training data to assess them confidently and because fair-lending and explainability rules require a defensible reason behind every declined application. The EU AI Act classifies creditworthiness assessment as a high-risk use case for exactly this reason, which pushes lenders towards models that can show their reasoning rather than treating the score as a black box.

The practical shift isn't underwriters being replaced. It's the underwriting queue narrowing to the cases that genuinely need a person.

Frequently asked questions

Can AI fully replace a human underwriter?

Not for every case. AI and decisioning engines now handle standard, well-documented applications end to end, but complex applications, thin credit files, and high-value loans still typically require a human underwriter's judgment, partly because models lack sufficient data on unusual cases and partly because regulation requires an explainable reason behind declined applications.

What credit score do you need to pass underwriting?

There's no universal minimum, since each lender sets its own risk appetite and weighs score alongside income, debt-to-income ratio, and the specific loan type. A lower score can still be approved with strong income and low existing debt, and a high score doesn't guarantee approval if affordability or documentation doesn't hold up.

How long does loan underwriting take?

An automated decisioning engine can underwrite a standard application in minutes. Manual underwriting, needed for complex or high-value cases, typically takes days rather than minutes, since it usually involves back-and-forth for additional documentation and a genuine human review of the file rather than a single automated pass.

Is underwriting the same in insurance and lending?

No. Loan underwriting assesses a borrower's ability to repay a loan. Insurance underwriting assesses the likelihood and cost of a future claim, using different risk factors entirely, such as health, property condition, or driving history rather than income and credit history. The word is shared; the process and risk being assessed are not.

Ready to automate more?

Put your customer operations on auto-pilot

Ready to automate more?

Put your customer operations on auto-pilot

Ready to automate more?

Put your customer operations on auto-pilot