AI underwriting: a buyer's guide to AI in insurance underwriting

What does AI actually do in insurance underwriting?

Two very different things are called AI underwriting, and it helps to keep them apart. The first is reading and checking: AI that opens a broker submission, an application, a schedule of values, a set of loss runs or a stack of medical records, keys the data, and checks it against appetite, guidelines and authority before an underwriter looks. The second is prediction: models that score risk or set price from data, often inside a rating engine. Both are AI in insurance underwriting, but they carry different risks and need different controls.

Most of an underwriter's week goes to the first kind of work, which is why it is where AI pays off soonest and with the least regulatory exposure. The underwriter still decides whether to quote, on what terms and at what price; AI makes sure the file in front of them is complete, accurate and traceable, with each figure linked to its page. Predictive models that change who is offered cover or what they pay sit squarely within insurance regulation and need testing, documentation and oversight from the start.

• Reading and checking: submission intake, clearance, data extraction, loss run analysis, guideline and authority checks, medical record summaries, wording comparison.
• Prediction: risk scores, propensity models, pricing factors and any model that uses external consumer data.
• Kept with people: decisions above an underwriter's authority limits or outside the rules your underwriters have approved, and the reasons given to an applicant. Rules-based binding within approved limits is used in personal lines, but people write, approve and monitor those rules.

Every insurance solution →The insurance buyer's guide →

How does AI handle commercial submission intake and clearance?

Commercial underwriting starts with an inbox. Brokers send ACORD applications, supplemental questionnaires, schedules of values, financials, driver lists and loss runs from prior carriers in every layout. Before anyone can price the account, someone has to read all of it, key it into the underwriting system, clear it against existing submissions, and check it against appetite. On a busy desk, the submissions that wait longest are often the ones a competitor quotes first.

AI underwriting takes that intake work end to end. It reads and keys each submission, clears it, turns prior-carrier loss runs into one loss view, checks every location's address, construction, occupancy, protection and value, and answers appetite and authority questions with the guideline that applies. For fleet and workers' compensation accounts it checks drivers and vehicles and rebuilds the experience mod. The underwriter opens a file that is ready to price and spends their time on the risk.

• Commercial Submission Intake & ClearanceAn intake solution for commercial carriers that reads, keys and clears every broker submission.
• Loss Run Analysis AutomationAn underwriting solution that turns prior-carrier loss runs in any layout into one loss view.
• Schedule of Values ValidationA property underwriting solution that checks every SOV location's address, COPE data and value.
• Underwriting Authority & Referral ManagementA guidelines solution that answers appetite and authority questions with the rule behind it.
• Commercial Auto Fleet UnderwritingA fleet underwriting solution that checks every driver, MVR and vehicle on an auto account.
• Experience Mod VerificationA workers' comp solution that rebuilds the mod from loss runs and payroll and explains gaps.

Commercial lines underwriting solutions →

Where does AI help personal lines underwriting?

Personal lines run on volume and rules. The underwriting questions are narrower, but they repeat across hundreds of thousands of policies: is the roof what the application says, has the garaging address or driver list changed, should binding stop in a county as a storm approaches, does this risk meet the eligibility guideline. The cost of getting them wrong shows up as premium leakage and as losses on risks that should have been priced differently.

AI helps by re-checking the book rather than a sample: reconciling applications with imagery and inspections, re-scoring renewals on new hazard and roof data, checking drivers, garaging and use after bind, and answering eligibility questions from the guidelines with the rule cited. Binding restrictions can be set and lifted by county and ZIP code with a record of who decided and when.

• Homeowners Renewal UnderwritingA renewal solution for home insurers that re-scores each policy on new hazard and roof data.
• Aerial Imagery VerificationA property underwriting solution that reconciles application, imagery and inspection findings.
• Auto Premium Leakage DetectionAn underwriting solution for auto insurers that checks drivers, garaging and use after bind.
• Binding Restriction ManagementA catastrophe underwriting solution that suspends and lifts binding by county and ZIP code.
• Underwriting Guidelines AssistantAn underwriting knowledge solution that answers eligibility questions from guidelines.

Personal lines underwriting solutions →

Can AI underwrite specialty, E&S and London market risks?

Specialty business is where underwriting judgment matters most and where documents are least standard: manuscript wordings, cyber control questionnaires and scans, towers of excess casualty policies, marine schedules that must be screened for sanctions, and surplus lines filings with their own diligent search and tax rules. AI does not replace the specialist here. It reads the wordings clause by clause against your standard form, checks each cyber control, assembles every underlying policy into one tower and shows the gaps, and screens vessels, owners and routes.

The value is speed with an audit trail. A specialist who no longer spends hours comparing wordings or chasing filings can see more submissions and write the ones that fit, and every difference they accepted is recorded with its source.

• Cyber Underwriting Risk AssessmentA cyber solution that checks each application, supplement and scan control by control for gaps.
• Policy Wording ComparisonA wordings solution that compares manuscript wordings clause by clause with your standard form.
• Excess Casualty Tower AnalysisAn excess underwriting solution that reads every underlying policy into one tower and its gaps.
• Marine Sanctions ScreeningA marine solution that screens each vessel, owner, route and cargo for sanctions and war risk.
• Surplus Lines Filing ComplianceA compliance solution for E&S placements that checks diligent search, tax and stamping filings.

Specialty, E&S and London market solutions →

How is AI used in life insurance underwriting?

Life and group underwriting depend on evidence about a person: the application, attending physician statements, lab results, prescription histories and evidence of insurability forms. Much of the delay comes from ordering and chasing those requirements and from reading long medical records. AI orders and tracks requirements, summarizes an attending physician statement into a cited summary an underwriter can check page by page, answers rating questions from the underwriting manual, and readies simple group evidence of insurability approvals for an underwriter to confirm.

This is also the area where regulators look hardest at models. The EU AI Act lists AI systems intended for risk assessment and pricing of natural persons in life and health insurance among its high-risk uses, and the first rule Colorado adopted under its 2021 law on insurers' use of external consumer data applied to life insurers. Keep the decision with the underwriter, and be ready to show what the model saw and why.

• APS SummarizationA medical review solution for life underwriters that turns APS pages into a cited summary.
• New Business Case ManagementA case management solution for life carriers that orders, chases and reads every requirement.
• Life Underwriting Manual SearchA guideline search solution for underwriters that answers rating questions from its manual.
• Evidence of Insurability AutomationAn EOI solution for group underwriters that reads each form and readies the simple approvals.

Life and annuities solutions →Group benefits and absence solutions →

How do carriers oversee underwriting done under delegated authority?

When an MGA or coverholder binds risks on a carrier's paper, the carrier is still responsible for what is written. Oversight usually means monthly bordereaux in each delegate's own format, binder terms that live in a contract, and periodic audits of a sample of risks. Problems such as a risk outside the binder's territory, a limit above authority or a class the binder excludes are often found months after they were bound.

AI makes oversight continuous. It maps and checks each bordereau as it arrives, tests every bound risk against its binder, answers an MGA underwriter's question about whether a risk fits with the clause cited, and audits delegates against the binder throughout the year rather than once. The capacity provider's team decides what to do about a breach; the system makes sure they see it.

• Bordereaux ManagementA bordereaux solution for carriers and MGAs that maps and checks each delegate's monthly file.
• Binding Authority ComplianceAn oversight solution for capacity providers that checks every bound risk against its binder.
• Binder Guidelines SearchA search solution for MGA underwriters that says if a risk fits the binder, citing the clause.
• MGA & Coverholder AuditAn audit solution for carriers that tests each delegate's risks and claims against the binder.

MGA and delegated authority solutions →

What do regulators expect of AI in underwriting and pricing?

The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted in December 2023 and adopted by 29 jurisdictions as of October 8, 2026, expects a written AI program covering the whole insurance life cycle, underwriting and pricing included, with governance, risk controls and internal audit, and it applies to systems bought from vendors as well as those built in-house. New York's Insurance Circular Letter No. 7 (July 2024) goes further for underwriting and pricing: insurers should not use AI systems or external consumer data unless they can show the use is not unfairly discriminatory, should test for disproportionate adverse effects and look for less discriminatory alternatives, may not rely solely on a vendor's claim of non-discrimination, and should give specific reasons for adverse decisions. Colorado's SB21-169, signed in July 2021, directs insurers, under rules the Division of Insurance adopts line by line, to test external consumer data, algorithms and predictive models for unfair discrimination and to correct it when found.

No software makes a model compliant by itself. What it can do is make the work of governance routine: keep an inventory of every model and where it is used, run and record unfair discrimination tests, show who pays more under a rate change before it is filed, and assemble rate and form filings. Ask any vendor how its product fits your AI program, and expect to test it like any other model.

• AI Governance & Model InventoryAn AI governance solution that lists every model and vendor AI tool, its tests and sign-off.
• AI Bias & Fairness TestingA bias testing solution that runs unfair discrimination tests on pricing and claims models.
• Rate Dislocation AnalysisA pricing solution for actuaries that shows who pays more under a rate change before filing.
• Rate & Form Filing AutomationA regulatory filing solution that assembles state filings and drafts objection responses.

Compliance, market conduct and AI governance solutions →How Botminds secures your data →

Does the same approach apply to lending credit decisions?

Credit underwriting at a bank or lender follows the same pattern: most of the effort is reading tax returns, statements and applications, and the decision belongs to a credit officer. The rules on explaining a decline are strict. Under Regulation B, the statement of reasons for an adverse action must be specific and indicate the principal reasons for it; saying only that the applicant did not meet the creditor's internal standards, or did not reach a qualifying score on its scoring system, is not enough, and the regulation makes no exception for decisions that rely on a complex model. The lessons carry over: let AI spread the financials and draft the memo with citations, keep the decision human, and be able to say exactly why.

• Commercial Financial SpreadingA spreading solution for credit teams that turns tax returns and audits into the bank's spread.
• Credit Memo AutomationA memo solution for credit analysts that drafts the full credit memo from the spread and file.
• Small Business Loan DecisioningAn underwriting solution for business banks that reads, spreads and tests small loan requests.
• Fair Lending Exception MonitoringA fair lending solution for compliance teams that compares pricing exceptions to similar loans.

The lending buyer's guide →

A checklist for any vendor

1. Which parts of your product read and check documents, and which predict risk or price?They need different controls. A good answer separates the two and explains how each is tested, documented and overseen.
2. Can it read our real submissions in the formats brokers actually send?Test on scanned loss runs, spreadsheets with merged cells, supplemental questionnaires and medical records, not on a vendor's demo set.
3. Does every appetite, guideline or authority answer cite the rule it applied?An underwriter, auditor or regulator must be able to see why a risk was cleared, referred or declined.
4. Who can bind, decline or change price, and what does the system record when they do?Look for named sign-off on every underwriting decision and an unalterable record of who decided, when and on what evidence.
5. What external consumer data does any model use, and how is it tested for unfair discrimination?Under NAIC, New York and Colorado expectations you remain responsible for vendor models; ask for the testing method, the metrics and how results are shared with you.
6. How will the product fit our written AI program and model inventory?Expect documentation you can file with your governance records, audit rights in the contract and cooperation with regulator questions.
7. How will we measure the pilot?Agree before you start: time from submission to quote, share of submissions answered, hit ratio, file accuracy against underwriter review, and override reasons.

Questions