How AI is changing commercial claims handling

Insurers use AI in commercial claims mainly for intake, document reading, summaries, triage, fraud screening and estimating. Complex commercial claims, such as a large fire with a business income loss or a serious liability suit, are still decided by people, and regulators expect that a person is genuinely reviewing any decision that hurts the policyholder. For a business, the practical effect is faster handling of small, simple claims and more reason to document your claim cleanly.

The evidence is thinner than the marketing. Most published numbers come from property or personal lines, or from vendors, and commercial lines are harder to standardize. Treat the figures below as indicators, not benchmarks.

Where AI shows up in a claim

Trade press describes the same uses. A 2025 Risk & Insurance feature quotes claims-service providers describing voicebots that take first notice by phone, models that flag claims likely to be litigated, tools that organize demand packages, and summaries of call transcripts. Straight-through processing is aimed at high-frequency, low-severity claims; see straight-through claims processing: what automated claims mean for your business.

How widely it is used

An Insurance Journal report in March 2026 on Sedgwick's property claims research put insurer use of AI tools at an estimated 58% to 82%, with 12% of carriers describing their capabilities as fully mature and 7% reporting scalable success. The same coverage reported that 75% of claims professionals believe AI needs human oversight. Reported gains cluster in intake and low-severity claims, where some carriers report 80% faster processing. The research does not isolate commercial claims.

That pattern fits the nature of commercial losses. A water-damaged restaurant with spoilage, equipment breakdown and a business income claim involves policy wording, forensic accounting and judgment. Software can sort the paperwork, but a person still has to decide what the policy covers.

What regulators expect

There is no federal rule on AI in claims. States regulate through insurance codes that already govern claims. The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers on December 4, 2023. According to the NAIC's implementation map (status as of October 8, 2026), 29 jurisdictions have adopted it: 27 states plus the District of Columbia and Puerto Rico. California, Colorado, New York and Texas have their own insurance-specific AI rules or guidance instead. See AI regulation in insurance: the NAIC bulletin and state rules.

The bulletin is guidance, not new law, and it is worth knowing what it says:

• AI used for "claim administration and payment" and fraud detection falls within the insurer's required written AI program, and the program should cover systems built by the insurer or bought from a vendor.
• Decisions made or supported by AI must still comply with the state's unfair trade practices and unfair claims settlement practices laws. The NAIC model claims act is written for claims under policies issued to residents of the state, not just consumers, which is why the same standards generally apply to commercial policyholders. Check your state's statute.
• In an exam or investigation, an insurer can expect to be asked about its governance, testing, third-party vendor oversight and the outcomes of its systems, including adverse outcomes for consumers.
• The bulletin ties the level of controls to the extent of human involvement in the final decision.

Some legislatures want more. Florida's House passed HB 527 unanimously in March 2026. It would have barred reducing or denying a claim solely on AI output and required a "qualified human professional" to review the facts and policy independently. The bill and its Senate companion died in committee, according to Repairer Driven News in April 2026. Expect the idea to return in Florida and elsewhere.

A hypothetical example

Say a small restaurant reports a kitchen fire through the carrier's app. The bot opens the claim, asks for photos, and triage sets it as moderate severity. An adjuster's estimate for the building arrives in a week, generated in part from the photos, and lists cleaning but not the damaged exhaust system behind the wall. The owner has the contractor and the hood-cleaning vendor each write a scope and invoice, submits them with the original photos, and points to the exact line the estimate omitted. The business income portion goes to a forensic accountant, to whom the owner supplies 24 months of financial statements and daily sales reports. The carrier revises the estimate. The facts are invented; the lesson is that software drafts the first number and a person can change it when handed better evidence.

How to work with an AI-assisted claim

1. Confirm intake in writing. Whatever channel you used, get the claim number, date and time of notice, and a summary of what you reported. A late-notice defense does not turn on whether a bot or a person took the call; see notice of claim and late notice: why timing matters.
2. Ask who the adjuster is. Get a named person with a phone number and email. Ask which person is responsible for coverage decisions.
3. Submit clean documents. Use PDFs with clear names, itemized invoices, dated photos with captions and one spreadsheet for amounts. Documents that software cannot read tend to be mis-extracted or ignored.
4. Check every summary and estimate. If a recorded statement transcript, scope of loss or letter misstates a fact, correct it in writing and ask that the correction be added to the file.
5. Do not accept "the system" as a reason. If a denial or partial payment is based on a policy provision, the letter should cite it. Ask for the specific provision, the facts relied on and the name of the person who made the decision. See what to do when your business insurance claim is denied.
6. Respond quickly to fraud flags. Most are false positives. Prompt, documented answers move the claim along; see commercial insurance fraud: premium fraud and staged claims.
7. Escalate with the right tool. A broker can raise the claim with the carrier's claims management. A public adjuster can handle the property and business income presentation. The state insurance department takes complaints about unfair claims handling; see how to complain to your state insurance department.

The business insurance claim checklist lists what to collect before you file.

When your own AI is the source of the loss

The other side of this topic is a claim arising from your use of AI, such as a chatbot that gives bad advice or an image tool that infringes. Verisk's ISO introduced generative AI exclusions for commercial general liability (CG 40 47, CG 40 48 and CG 35 08, January 2026 edition), reported by Insurance Business in September 2026, and carriers file their own wording. Check what your policies say before relying on them. See AI liability insurance: how insurers are treating artificial intelligence risk.

Key takeaways

• AI is a tool in intake, triage, document reading, estimating and fraud screening; people still decide the complicated commercial claims.
• Industry data on AI claims handling is mostly vendor-reported and not specific to commercial lines, so read the numbers cautiously.
• State claims laws apply whether or not software helped; the NAIC bulletin (adopted by 29 jurisdictions as of October 8, 2026) tells insurers to keep AI under governance and human review where decisions affect policyholders.
• Document your claim for machine and human readers, confirm intake in writing, challenge estimates with your own experts and ask for the policy provision behind any denial.
• If an AI-assisted decision looks wrong, escalate through the broker, a public adjuster or counsel, and the state insurance department.