AI in Insurance: Where It Is Used Today, by Function

On this page (15)

1. TL;DR
2. AI in insurance at a glance
3. How widely insurers use AI
4. Distribution and agencies
5. Underwriting
6. Pricing and rating
7. Policy administration and core systems
8. Claims: FNOL, estimating and fraud
9. Document intake
10. Customer service
11. Compliance and AI governance
12. Regulation of AI in insurance
13. AI insurance companies: carriers vs vendors
14. How to pick where to start
15. FAQs

TL;DR

• AI in insurance is concentrated in claims, underwriting, marketing and fraud detection. In the NAIC's homeowners survey, 54% of responding insurers reported AI/ML use in claims, 47% in underwriting and in marketing, 42% in fraud detection and 35% in rating.
• Adoption differs by line. The share of insurers that use, plan to use or plan to explore AI/ML was 88% in private passenger auto, 70% in homeowners, 58% in life and 92% in health, across NAIC surveys published from 2022 to 2025 (the surveys define AI differently, so compare with care).
• Most AI that insurers run is bought from vendors: core systems (Guidewire, Duck Creek), point tools for one job (Tractable for damage photos, Shift Technology for fraud) and general assistants. The homeowners survey found 42% of reported models came from a third party.
• Regulation is state by state. The NAIC's map shows 29 jurisdictions had adopted its December 2023 AI Model Bulletin as of October 8, 2026, and California, Colorado, New York and Texas have their own insurance AI guidance.
• AI insurance companies means two different things: insurtech carriers that sell policies (Lemonade, Root) and software vendors that sell AI to insurers and agencies. Buyers evaluating tools want the second group.

AI in insurance is used most in claims, underwriting, marketing and fraud detection, and least in loss prevention. In the NAIC’s survey of homeowners insurers, 54% of the 194 responding companies reported using AI or machine learning in claims, 47% in underwriting, 47% in marketing, 42% in fraud detection, 35% in rating and 14% in loss prevention. The concrete jobs are narrower than the headlines suggest: sorting incoming submissions, reading documents, estimating damage from photos, referring suspicious claims to investigators and drafting answers for service staff. Most of it is bought from vendors rather than built in house.

This page walks the insurance value chain one function at a time. For each it covers what AI does, the kinds of vendors that sell it, and links to the comparison list for that category.

AI in insurance at a glance

Function	What AI does there	Example vendors	Compare
Distribution and agencies	Quoting, CRM follow-up, email and summary drafting, data entry from carrier documents	Applied Epic, Vertafore, InsuredMine	AI tools for agencies
Underwriting	Submission intake, appetite checks, triage and scoring, risk data enrichment	Sixfold, Kalepa, Cytora	Underwriting workbench
Pricing and rating	Machine learning and GLM/GAM rate modeling, demand modeling, rating engines	Akur8, Earnix	Rating and pricing
Policy administration	AI features layered onto core policy, billing and claims systems	Guidewire, Duck Creek, Socotra	Policy administration
Claims	FNOL intake, triage, photo estimating, total loss, subrogation	CCC, Tractable, Five Sigma	Claims management, Claims estimating
Fraud	Scoring claims and applications, network links, SIU referral	Shift Technology, FRISS	Fraud detection
Document intake	Extracting fields from ACORD forms, loss runs, SOVs and policies	Hyperscience, Indico Data, Sortspoke	Document extraction
Customer service	Chat and text intake, message suggestions, translation, drafting	Hi Marley, general AI assistants	AI assistants
Compliance and governance	Model inventory, bias testing, monitoring, audit records	Monitaur, Credo AI	AI governance

How widely insurers use AI

The best public adoption data comes from the National Association of Insurance Commissioners (NAIC), which has surveyed insurers line by line under state market examination authority. Its artificial intelligence page reports these results:

Line of business	Report issued	Insurers responding	Use, plan to use or plan to explore AI/ML
Private passenger auto	December 2022	193	88%
Homeowners	August 2023	194	70%
Life	December 2023	161	58%
Health	May 2025	93	92%

Two cautions before quoting these numbers. The auto and homeowners surveys counted only “more advanced” models and excluded generalized linear models (GLMs) and GAMs, while the life and health surveys included them, so the health figure is not directly comparable with the homeowners one. The NAIC also counts plans and exploration, not only production use. For health, the NAIC’s May 2025 release puts current use at 84% of responding insurers.

The homeowners memo has the most useful split by function, and it also shows where models come from. Of 2,413 reported model uses, 42% were developed by a third party and 58% in house, with 95 distinct third-party model vendors named. Marketing had the most third-party vendors (48), followed by claims (37). The NAIC says the auto survey’s in-house versus vendor split was nearly identical.

Distribution and agencies

For agencies and brokers, AI shows up as features inside the systems they already pay for and as general assistants on the side. Typical jobs:

• Pre-filling applications and quoting several carriers from one entry. Comparative raters such as EZLynx do this for personal lines, and Bold Penguin does it for small commercial.
• CRM work: renewal reminders, cross-sell prompts and follow-up sequences, as in InsuredMine and AgencyZoom.
• Drafting client emails, coverage summaries and meeting notes with assistants such as ChatGPT, Claude or Microsoft Copilot. The policy summary prompts page has templates for this.
• Moving carrier downloads and documents into the agency management system.

The large agency management vendors are adding AI to their platforms. Vertafore introduced its Velocity AI Platform and a first set of AI agents in April 2026, running inside its existing product families, and Applied Systems acquired the AI data company Planck in July 2024. Compare the systems on agency management systems, the CRMs on insurance CRM software, and the AI-specific options on AI tools for insurance agencies. Published pricing for two common systems is broken down in the EZLynx pricing and HawkSoft pricing guides.

Underwriting

Commercial underwriting runs on submissions: an email with an ACORD application, loss runs, a statement of values and a broker’s notes. Most of the AI investment here goes to reading that pile and deciding which submissions deserve an underwriter’s time.

• Intake and extraction. Software pulls the insured’s details, exposures and loss history out of the attachments into a structured record.
• Appetite and guideline checks. Sixfold says it scores each risk with explainable signals and cites every fact to its source document. Kalepa describes appetite guardrails and loss run analysis for specialty lines.
• Triage and prioritization. Federato pitches portfolio-aware ranking, so underwriters see which risks fit the book’s current concentration.
• Data enrichment. Vendors such as Planck build a business profile from a name and address, and property data providers score roofs and hazards from aerial imagery (see property data providers, including ZestyAI and Cape Analytics).

The NAIC homeowners memo found that most underwriting models were developed in house and that many were used for automation rather than support. Head-to-heads: Kalepa vs Sixfold and Federato vs Kalepa. Full list: underwriting workbench software.

Pricing and rating

Actuarial pricing was using statistical models long before the current AI wave, which is why the NAIC excluded GLMs from two of its surveys. The newer tools speed up model building and add machine learning while keeping the output explainable enough to file with regulators. Akur8 says its algorithms produce transparent GLMs and GAMs and that its rating engine exposes the plan as an API. Earnix connects pricing models, demand and elasticity to production rating. Core vendors sell their own rating modules, including Guidewire PricingCenter and Duck Creek Rating.

In the homeowners survey, 35% of insurers reported AI/ML use in rating, the second-lowest function. Compare options on insurance rating and pricing software.

Policy administration and core systems

Policy administration systems are the carrier’s system of record for products, policies, billing and often claims. These are the “insurance software companies” most carriers mean: Guidewire, Duck Creek, Majesco, Sapiens, EIS, Insurity and API-first platforms such as Socotra. AI here mostly arrives as features and partner apps (document ingestion, assistants for underwriters and adjusters, analytics) rather than as the reason to buy the platform. A core replacement takes years, so carriers more often add AI tools around the core through APIs.

Compare on policy administration systems and Duck Creek vs Guidewire.

Claims: FNOL, estimating and fraud

Claims is where insurers report the most AI use. The NAIC homeowners memo lists subrogation and claims triage as the top uses (44 companies each), followed by evaluating images of loss (36). Only 2 companies reported using AI/ML to determine the settlement amount, and over 60% of the triage and subrogation models were reported as augmenting human decisions.

First notice of loss (FNOL). Chat, text and app intake capture the loss and route it. Hi Marley runs claims conversations over SMS, and Snapsheet and Five Sigma automate steps from intake onward. Five Sigma says its AI adjuster, Clive, can run on top of an existing claims system. Platforms are compared on claims management software.

Estimating. In auto, photo-based estimating is the most mature use of computer vision in insurance. CCC, Mitchell, Tractable, Solera Qapter and Claim Genius turn damage photos into estimates or flag estimates for review. See AI claims estimating software and CCC vs Tractable. Adjusters drafting reports with general assistants can start from the adjuster report prompt library.

Fraud. In the homeowners survey, 42% of insurers reported AI/ML in fraud detection, mostly to refer claims for further investigation, and fraud was the area where models came mostly from third parties. Vendors score claims and applications and link related parties: Shift Technology, FRISS, Verisk ClaimSearch and SAS. Compare on insurance fraud detection software and FRISS vs Shift Technology.

Document intake

Insurance moves on PDFs: applications, loss runs, schedules, policies, invoices and medical records. Extraction software reads them and returns structured fields, which feeds underwriting, policy checking and claims. The layouts vary by carrier, which is why insurance-specific tools ship with ACORD and loss run models. Vendors range from enterprise intelligent document processing (Hyperscience, Indico Data) to insurance-native tools (Sortspoke, Roots) and developer APIs (Reducto).

Lists by document type: insurance document extraction, ACORD form extraction, loss run parsing and SOV extraction. Background on the most common commercial document is in what is a loss run report.

Customer service

Carriers and agencies use AI to answer routine questions, take changes and payments, translate messages and draft replies for staff. Lemonade, a carrier, says its bot AI Maya handles onboarding from collecting information to creating quotes. Most agencies use general assistants for drafting rather than customer-facing bots; the four main ones are compared on AI assistants for insurance and ChatGPT vs Claude. Templates for appeals and denial letters are on insurance claims denial prompts.

Compliance and AI governance

Regulators expect insurers to know which AI systems they run, what data feeds them and how they were tested. Governance software keeps a model inventory, runs bias and performance tests, monitors models in production and stores the records an examiner might ask for. Vendors include Monitaur, Credo AI, ValidMind, IBM watsonx.governance, Holistic AI and Arthur. Compare on AI governance software for insurers and Credo AI vs Monitaur.

In the 2025 health survey, the NAIC reports that nearly 92% of responding health insurers had AI/ML governance principles modeled on the NAIC AI Principles.

Regulation of AI in insurance

Insurance is regulated by the states, and AI rules follow that pattern.

• NAIC Model Bulletin. The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers on December 4, 2023. It expects insurers to keep a written AI program, manage risk and internal controls, oversee third-party AI systems and data, and answer regulators’ questions about their AI use. It has no effect in a state until that state issues it.
• Adoption. The NAIC’s implementation map, status as of October 8, 2026, lists 29 adopting jurisdictions: Alaska, Arkansas, Connecticut, Delaware, the District of Columbia, Hawaii, Illinois, Iowa, Kentucky, Maryland, Massachusetts, Michigan, Mississippi, Nebraska, Nevada, New Hampshire, New Jersey, North Carolina, North Dakota, Oklahoma, Pennsylvania, Puerto Rico, Rhode Island, South Carolina, Vermont, Virginia, Washington, West Virginia and Wisconsin. States can modify the text when they adopt it.
• Own rules. The same map lists insurance-specific AI guidance in California (Bulletin 2022-5), Colorado (3 CCR 702-10), New York (Circular Letter No. 7, July 2024) and Texas (Bulletin B-0036-20).
• Colorado. SB21-169, signed July 6, 2021, requires insurers to test external consumer data, algorithms and predictive models for unfair discrimination and to correct harms they find. Amended Regulation 10-1-1 sets governance requirements for life, private passenger auto and health benefit plan insurers, and insurers that do not use such data must file an officer-signed attestation by December 1 each year. The Division has also circulated a draft quantitative testing regulation for life underwriting. Colorado’s separate general AI statute, SB 24-205, and its 2026 changes are covered in Colorado SB 24-205 and insurance AI.
• Next steps at the NAIC. The NAIC says 12 states were piloting an AI Systems Evaluation Tool as of March 2026, with adoption anticipated at its 2026 Fall National Meeting.

This is a summary of public sources, not legal advice.

AI insurance companies: carriers vs vendors

Searches for “AI insurance companies” mix two kinds of business.

Insurtech carriers hold insurance licenses, take risk (often with reinsurance) and use software and data across their own operations.

• Lemonade describes itself in its 2025 annual report as a vertically integrated company with wholly owned insurance carriers in the US and Europe. It says its bot AI Maya and its APIs sell 98% of its policies; as of December 31, 2025, its claims bot AI Jim took the first notice of loss without human intervention 96% of the time, and roughly 55% of claims were automated from start to finish. Claims AI Jim is not authorized to settle go to human adjusters.
• Root calls itself a technology insurance company built on the idea that car insurance rates should rest mainly on driving behavior, using telematics and data science to price, per its 2025 annual report.
• Clearcover sells direct auto insurance priced with its own algorithms and serviced through a mobile app.

AI vendors sell to insurers, MGAs and agencies and do not carry risk. Almost all the tools linked on this page fall in this group. If you work at a carrier or agency and are shopping for software, these are the companies to evaluate, starting from the list for the function you want to change.

How to pick where to start

Start with the function where staff spend the most time re-keying or reading: for most commercial carriers and MGAs that is submission intake, and for agencies it is data entry and client communication. Ask each vendor to run your own documents or claims, ask what the model decides versus what it suggests to a person, and ask for the documentation your state’s version of the NAIC bulletin expects you to keep on third-party AI.

FAQs

How is AI used in the insurance industry?

Insurers use AI to sort and score incoming submissions, set prices with machine learning models, take first notice of loss, estimate damage from photos, flag suspicious claims, pull data out of documents and answer customer questions. NAIC surveys of homeowners insurers found the highest reported use in claims (54% of respondents), followed by underwriting and marketing (47% each).

What are AI insurance companies?

The phrase covers two groups. Insurtech carriers such as Lemonade and Root are licensed insurers that built their operations around software and data. AI vendors such as Shift Technology, Tractable or Sixfold do not sell insurance; they sell software that carriers, MGAs and agencies use in claims, fraud or underwriting.

Which insurance software companies are the biggest?

No public, neutral ranking settles this. For carriers, the core policy, billing and claims platforms most often shortlisted include Guidewire, Duck Creek, Majesco, Sapiens, EIS and Insurity. For agencies, Applied Systems and Vertafore are the largest agency management vendors. See the policy administration and agency management lists for comparisons.

What insurance company software uses AI?

Core platforms are adding AI features, but most specialized AI comes from add-on tools: underwriting workbenches (Sixfold, Kalepa, Cytora), pricing platforms (Akur8, Earnix), claims estimating (CCC, Tractable, Mitchell), fraud detection (Shift Technology, FRISS) and document extraction. Each has its own list on this site.

Is AI in insurance regulated?

Yes, mainly at the state level. The NAIC adopted a Model Bulletin on insurers' use of AI systems in December 2023, and 29 jurisdictions had adopted it as of October 2026 according to the NAIC's map. Colorado regulates life, private passenger auto and health insurers' use of external data and predictive models under SB21-169 and Regulation 10-1-1.

Does AI decide insurance claims on its own?

Sometimes, for simple claims. Lemonade's 2025 annual report says roughly 55% of its claims were automated end to end, with the rest passed to human adjusters. In the NAIC homeowners survey, only 2 companies reported using AI/ML claims models to determine settlement amounts, and most claims models were reported as supporting human decisions.

Will AI replace insurance agents?

The evidence so far points to AI taking over tasks inside agencies (data entry, quoting, renewal follow-up, email drafting) rather than replacing the agent channel. Lemonade says its bot and APIs sell 98% of its policies, but it notes that US homeowners insurance is sold primarily through agents.