I was listening to the debate between Steve Badger and René Sigman at the NAPIA annual meeting in Dallas yesterday when Badger mentioned the new AI Bulletin from the Texas Department of Insurance. That caught my attention. Any time an insurance regulator uses the words “artificial intelligence” and “claims” in the same breath, claims professionals should pay attention.
The Texas Department of Insurance issued Commissioner’s Bulletin B-0003-26 on June 12, 2026. Its subject is simple: “Use of artificial intelligence.” Texas is telling insurers and all regulated entities that AI is not some magic loophole allowing them to do indirectly what the law prohibits them from doing directly.
The bulletin says that decisions or actions impacting consumers, when made or supported by advanced analytical and computational technologies, including AI, must comply with all applicable insurance laws and regulations. This includes laws addressing unfair trade practices and unfair discrimination. In other words, a carrier cannot shrug its corporate shoulders and say, “The computer did it.”
For those of us in the property insurance claim world, the key part is TDI’s specific reference to Texas Insurance Code Chapter 542, which governs the processing and settlement of claims. The chapter sets standards for the investigation and disposition of claims. The bulletin also references Chapter 541 regarding unfair trade practices, Chapter 544 regarding prohibited discrimination, and Chapter 4101 regarding adjuster licensing. The last point matters greatly. Texas law requires licensed adjusters for certain acts, including investigating or adjusting losses and supervising the handling of claims.
An algorithm may sort data. It may flag patterns. It may estimate. It may compare photographs, Xactimate line items, prior claims, weather reports, or internal claim notes. But an algorithm is not a licensed adjuster. It does not owe a duty of good faith. It cannot sit for an examination under oath. It cannot be cross-examined about why it ignored hail hits on a metal roof, downgraded labor hours, or quietly treated a full replacement claim as a patch job.
TDI’s bulletin requires that decisions made using AI must not be inaccurate, arbitrary, capricious, or unfairly discriminatory. Those are strong words. The bulletin further states that if a regulated entity uses AI to make a consequential decision, TDI expects a person to review and agree with all decisions before action is taken. This is a very practical requirement. If an insurer denies a claim, reduces an estimate, flags fraud, delays payment, or makes a coverage determination with the assistance of AI, who reviewed it? What did that person review? Did the human being merely rubber-stamp the machine? Was that person licensed? Was that person trained? Did that person understand the policy, the facts, and the applicable Texas claims handling statutes?
For years, I have warned that claims departments are increasingly designed around speed, metrics, leakage control, and vendor-driven uniformity. Artificial intelligence can make that problem worse if it becomes a faster way to underpay claims at scale. A bad claim decision made by one poorly trained adjuster harms one policyholder. A bad claim decision embedded in an AI model can harm thousands before anybody realizes the machine has been steering the ship toward the rocks.
The bulletin also tells insurers to have governance, risk management, internal controls, verification, testing, and audit functions. TDI says it may ask about governance frameworks, data and privacy protections, internal controls, and any specific use of AI or its application. That should be music to the ears of those representing policyholders. The black box is not supposed to stay black when a regulator comes knocking. It should not stay black in litigation either.
This bulletin is not an anti-technology manifesto. Technology can help claims handling. Used properly, AI may help identify damage, speed communications, detect inconsistencies, and assist adjusters with better information. But technology must serve the insurance promise. It cannot replace independent judgment, licensed adjusting, fair investigation, and prompt payment of what is owed.
For public adjusters, the lesson is straightforward. Ask whether AI was used. Ask what tool was used. Ask whether a third-party vendor was involved. Ask what data was fed into the system. Ask whether the output was reviewed by a licensed adjuster. Ask whether the insurer tested the system for error and bias. Ask whether the claim file contains the AI output, prompts, scoring, flags, recommendations, or override history.
For policyholder lawyers, the discovery roadmap is now brighter. Requests for production should include AI governance materials, vendor contracts, model documentation, audit results, training materials, claim-specific outputs, adjuster review notes, and communications about the use of AI in the claim. Depositions should ask whether the so-called human review was meaningful or merely ceremonial.
Texas has a reputation for independence. Maybe that is why this bulletin matters. It tells insurers operating in Texas that accountability cannot be outsourced to a machine. The insurance company sold the policy. The insurance company accepted the premium. The insurance company owes the claim handling duties and remains responsible for the result.
Thought For The Day
“You may all go to hell and I will go to Texas.”
— Davy Crockett