Does a customer have to be told an AI system was involved in their quote or claim decision?
In the states adopting the NAIC model bulletin, the working answer is yes. The harder part is proving your process: what the AI did, who reviewed it, and what the customer was told.
Answered in short
5 things that decide this
- 01In states that adopted the NAIC model bulletin on insurers' use of AI, roughly 25 as of mid-2026, customers should be told when an AI system played a part in a decision about their insurance.
- 02Carriers were the bulletin's first audience. Agencies that use AI-assisted quoting or triage are held to the same rules, and nobody hands an agency a translated checklist.
- 03State wording varies. California, Colorado, New York and Texas run their own AI frameworks rather than the model bulletin, so the duty to tell depends on where the customer sits.
- 04Telling the customer is not enough on its own. Your record must show what the tool touched, who reviewed its output before it went out, and when.
- 05A 12-state pilot of an AI Systems Evaluation Tool runs through September 2026, a sign that examiners are building the muscle to check this.
What the NAIC bulletin expects when AI touches a decision
The NAIC model bulletin sets expectations for insurers that use AI in decisions affecting consumers. Telling the customer is one of them. A documented program is the other: proof the systems are governed and their outputs reviewed. About 25 states had adopted it by mid-2026. It reaches quoting and claims alike, and a market-conduct examiner can ask for the records behind it.
For an agency, the awkward part is translation. Its language was written for carriers. Yet if your agency runs AI-assisted quoting or triage, you own a slice of the same duties: a notice process for your customers, and a record showing a person reviewed what the tool suggested.
Four states went their own way. California, Colorado, New York and Texas run separate AI frameworks, so a multi-state book can face different notification wording for the same quote.
- Quote assistedThe tool drafts or recommends.
- Human reviewA licensed person checks before it goes out.
- Customer notifiedPlain-language notice that AI helped.
- LoggedWhat ran, who reviewed, when.
- Exam-readyRecords available on request.
An examiner asks two questions: did the customer know, and did a person check.
What an agency should put in place
You can cover the duty with three pieces. First, a customer-notice template matched to your state's adoption of the bulletin. Second, a required human check before any AI-assisted suggestion reaches a client. Third, a log of every AI-assisted quote or coverage suggestion. Together they answer the two questions an examiner asks: did the customer know, and did a person check.
The record-keeping shape is one we've built before. PremiumAudit.io, an insurance audit platform Hashlogics built, moved carriers, auditors and policyholders onto one shared record instead of traded spreadsheets. An AI-use log for an agency is the same discipline pointed at compliance.
Keep the notice itself honest and short. A plain sentence telling the customer an automated system helped, with a person staying accountable, does the job better than a legal wall of text.
Related questions
01Does this cover claims as well as quotes?+
Yes. The bulletin covers AI used in decisions that affect consumers, which reaches quoting and claims handling alike. Wording varies with each state's adoption, so check your state first.
02We only use the AI features inside our rater. Does that count?+
For practical purposes, yes. If an automated suggestion shapes what a customer is quoted or told, the review duty is yours even though the tool isn't. Log what the feature does. Keep a person in the loop before its output reaches a client.
03What happens if we say nothing?+
In an adopting state, saying nothing is a bet that no complaint, audit or market-conduct exam ever asks about your AI use. The notification costs a sentence and the record is cheap to keep, next to explaining its absence. It's a bad bet.
Related
- Software for insurance agencies →Our insurance hub: what agencies actually buy.
- AI agent development →AI systems built with the review checkpoint and the log in place.
- Insurance AI wins where the rules are written down →Why regulated workflows are where insurance AI pays off.
- How to automate loss-run and ACORD extraction →The document side of the same submission flow.
