AI is changing insurance work across underwriting, pricing, claims, policy issuance and health-plan administration. But survey figures that combine current use with plans to adopt or explore AI do not show how many insurers have systems in production. In the United States, National Association of Insurance Commissioners (NAIC) summaries show reported interest across auto, home, life and health insurance, while its model guidance highlights the need to manage risks when systems affect consumers.
What the insurance AI adoption figures do—and don’t—show
The NAIC’s published summaries report the share of responding insurers that said they use, plan to use, or plan to explore artificial intelligence or machine learning (AI/ML). Those categories are combined: the figures are not deployment rates, and they do not establish how often systems are used or what results they produce. The surveys also cover different insurance lines and were issued in different years, so they should not be read as a trend measured on a common basis.
As an Amazon Associate I earn from qualifying purchases.
| Insurance line | Respondents | Reported use, plans to use, or plans to explore AI/ML | NAIC aggregate report |
|---|---|---|---|
| Auto | 193 | 88% | December 2022 |
| Home | 194 | 70% | August 2023 |
| Life | 161 | 58% | December 2023 |
| Health | 93 | 92% | May 2025 |
These are findings about the survey respondents, not a census of every insurer in the United States. They also do not measure AI’s effect on premiums, claim speed, accuracy, costs or consumer outcomes. A high combined figure can reflect exploration or planning as well as active use.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Where AI can enter the insurance workflow
Insurance is not one uniform AI use case. A model that flags an application for review has a different role from one that estimates a claim or contributes to a price. The NAIC’s summaries describe applications across the policy lifecycle, including marketing, underwriting, pricing, servicing and claims.
#1 Best Overall
Auto and home: pricing, inspections and claims
- Risk scoring and rate factors: Machine-learning models may help assess risk or estimate the relativities used in pricing. The NAIC reports that auto and home insurers mostly developed pricing and underwriting models in-house; that observation applies to those lines and model categories, not every insurer or AI system.
- Renewals and inspections: AI/ML may support renewal evaluation and inspections, including assessing information about a property or policyholder.
- Accident-image analysis: Image analysis can contribute to evaluating damage after an accident.
- Claim estimates and fraud detection: Models may estimate ultimate claim settlement values or flag patterns for possible fraud review. A flag or estimate is an input to a process; it does not, by itself, establish fraud or determine the final settlement.
Life: offers, underwriting and policy issuance
Reported life-insurance applications include targeted offers, shortening policy issuance, supporting approval or denial decisions, and assigning underwriting risk classes. These uses can influence whether an application proceeds and how it is classified, so the consequences depend on how the model’s output is used in the insurer’s decision process.
Health: authorization, claims and plan operations
Health-insurance examples include prior authorization, claims adjudication, fraud detection, risk adjustment, pricing and plan design, processing, and sales and marketing. These are distinct tasks: a system used for administrative processing may not have the same consumer impact as one that informs coverage or payment decisions.
Marketing and third-party models
The NAIC summary says roughly half of models used for marketing were developed by third-party vendors. That is a specific observation about marketing models, not a general estimate for all insurance AI. Vendor involvement matters because an insurer may need to understand and oversee a system it did not build itself.
What it means when a model supports a decision
AI systems may automate a decision, augment a person’s judgment, or provide information that supports a decision. Those roles should not be treated as interchangeable. An automated outcome may leave little room for a person to catch a bad input or an unsuitable result before it affects a consumer. A human reviewer can add scrutiny, but only if the reviewer has enough information, authority and time to question the model rather than simply accept its output.
When assessing an AI-supported process, insurers and regulators need to look beyond whether a human is nominally involved. The relevant questions include what the model can decide, what the reviewer can change, and what happens when information is incomplete or a consumer disputes an outcome. The more consequential and difficult to reverse a decision is, the more important it is to understand how errors are detected and addressed.
Potential efficiency gains come with consumer and operational risks
The NAIC’s December 2023 Model Bulletin recognizes that AI may support innovation, improve consumer interfaces, simplify processes, and contribute to efficiency or accuracy. These are potential benefits, not proof that a particular insurer’s system achieves them. The NAIC also identifies risks that can affect consumers and insurers:
Rank #3
- Inaccuracy: Poor data, unsuitable assumptions or model errors can produce unreliable classifications, estimates or recommendations.
- Unfair discrimination: A model or its data may produce outcomes that conflict with applicable laws, even when the system does not explicitly use a protected characteristic.
- Data vulnerability: Sensitive insurance information can be exposed or mishandled in the systems and processes used to develop or operate models.
- Limited transparency or explainability: Insurers may struggle to understand, document or explain why a system produced an output, particularly when relying on a vendor’s model.
These concerns are connected. For example, weak data controls can undermine model reliability, while limited documentation can make it harder to investigate a questionable result. Efficiency should therefore be evaluated alongside decision quality, fairness, security and the ability to explain and correct errors.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWhat the NAIC Model Bulletin means for insurers
The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers in December 2023. It reminds insurers that consumer-impacting decisions or actions made or supported by advanced analytical and computational technologies must comply with applicable insurance laws, including rules addressing unfair trade practices and unfair discrimination. It also describes governance expectations and information regulators may request during an investigation or examination.
The bulletin is model guidance, not a single nationwide statute. Its legal effect depends on state action, and insurers remain subject to other applicable laws. The NAIC’s 2020 AI principles, as summarized on its official pages, emphasize fairness and ethical use, accountability, compliance, transparency, and safety, security, fairness and robustness.
For insurers, the practical implication is that adding a model does not remove responsibility for the consumer-facing decision. Governance needs to cover systems developed internally as well as those supplied by vendors, and it needs to account for how a tool is actually used in a particular workflow.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Regulatory evaluation is evolving
As of the NAIC topic page’s April 3, 2026 update, the Big Data and Artificial Intelligence Working Group was developing an AI Systems Evaluation Tool for regulatory examinations and analyses. The page reported that a pilot involving 12 participating states was underway in March 2026 and that adoption was anticipated at the NAIC’s 2026 Fall National Meeting. That was an anticipated milestone, not confirmation that adoption occurred. The working group’s 2026 charge includes researching insurer AI, monitoring regulatory developments and facilitating evaluation of AI systems.
Free tools Windows power users keep installed
One-click scans. No signup required.
This work signals an effort to make regulatory review more systematic; it does not establish a single evaluation procedure or legal standard for every state. Insurers should distinguish NAIC activity from requirements actually adopted in the jurisdictions where they operate.
Best Value
Questions to ask before relying on an insurance AI system
A useful review follows the model from its inputs to the consumer outcome. The answers should be specific to the system, workflow and jurisdictions involved.
- What decision or action does it influence? Identify whether it affects marketing, pricing, underwriting, renewal, authorization, claim handling or another process, and whether it recommends, ranks, flags or makes a decision.
- What data does it use, and where did that data come from? Check data quality, relevance, permissions, security and whether the information remains appropriate for the intended use.
- Who developed and maintains it? Establish what the insurer can inspect and document when a model or key component comes from a vendor.
- How is performance validated? Set out how the insurer tests reliability and looks for inaccurate or unfair outcomes before use and as conditions change.
- How is it monitored after launch? Define who reviews results, what signals trigger investigation, and how the insurer can pause, correct or replace a system that is not performing as intended.
- What can a human reviewer do? Clarify when review is required, what information reviewers receive, and whether they can meaningfully override or escalate an output.
- How can an affected consumer get an understandable answer or correction? Consider what explanation is available, how disputed information is checked, and how errors can be remedied in the relevant process.
The answers reveal more than an adoption label. They show whether an insurer can account for a system’s role, oversee its risks and explain the decisions that affect consumers.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
