Will AI Replace Insurance Agents? What the Job Data Shows
Federal data rates insurance agents among the jobs most exposed to AI and still projects their numbers to grow through 2035.
- For
- Insurance agents, producers, agency principals, and distribution leaders.
- Read if
- You sell insurance or run an agency, and you want to know what the federal job data says about AI replacing agents and which rules reach your own use of AI.
In this article
On August 27, 2026, the federal Bureau of Labor Statistics (BLS) published its employment projections for 2025 to 2035, with a new ranking of occupations by how much of their work AI could do or help with.1 Insurance sales agents landed in its top group, “Very high.” The projections still expect agent numbers to grow.2 One finding measures the work and the other counts the agents.
What the federal projection counts
The projection expects agent jobs to grow 3 percent over the decade, about as fast as the average for all occupations: a net gain of roughly 18,800 on a base of about 572,600.2 These are national figures, not forecasts for any state, line of business or agency.
Openings, the jobs employers must fill each year whether new or vacated, come to about 43,100 for agents, many left by workers who change occupations or leave the labor force. That is more than twice the decade’s net gain, so the figure measures hiring, not growth.
According to a 2025 BLS article, the projection methods assume technology keeps changing at its past pace and are not designed for extremely rapid change. They generally build in a technology’s effects only when BLS research gives clear expectations, and many effects of generative AI are too uncertain to include.3 For sales occupations overall, the August release expects e-commerce and AI tools in selling to keep reducing jobs.1 For agents, BLS’s page names only online research and buying as cutting demand and never mentions AI.2 That leaves open what the models weighed.
The case for replacement
In 2013, before large language models, Carl Benedikt Frey and Michael Osborne estimated the probability that each occupation’s work could be automated by computer-controlled equipment.4 Insurance sales agents scored 0.92, well inside the group they called high risk and potentially automatable “over some unspecified number of years, perhaps a decade or two.”4
Lemonade, an insurance company, says in its 2025 annual report that its chatbot, AI Maya, and its APIs sell 98 percent of its policies. Maya collects a customer’s information, personalizes coverage, creates quotes and handles payment.5
Neither measures jobs. Frey and Osborne made no attempt to estimate how many would actually be automated,4 and thirteen years into their decade or two, BLS still expects the occupation to grow. Lemonade calls a platform like its own an outlier in US homeowners insurance, which it says is sold primarily through agents.5
Why BLS still expects agents to grow
BLS cautions that a “Very high” rating does not necessarily mean employment will decline.1 For agents, it gives a reason: they will still be needed to help clients understand their options and choose a policy.2
Three other insurance occupations rated “Very high” are expected to shrink: underwriters, claims adjusters and examiners, and the clerks who process claims and policies. Technology is among the reasons for each, AI included for claims workers.6
On our reading, the difference is between processing and advice. The shrinking insurance jobs move applications, claims and policies through steps that software speeds up; agents’ growth rests on clients who want help deciding.
What agencies use AI for
Among independent agencies, AI use rose from 15 percent in 2024 to 46 percent, according to the 2026 Agency Universe Study by the Independent Insurance Agents & Brokers of America, run with a coalition that includes carriers. Uses include marketing content, coverage form analysis and contract review. Agencies were comfortable letting customers serve themselves for policy documents, billing and claims, but not for quotes or purchases. Staff per agency averaged 9.9 across all positions, up from 8.2 in 2024, though the published findings do not tie the rise to AI.7
Two of those answers suggest AI in these agencies is helping the adviser rather than replacing one: coverage form analysis is close to the help with options BLS cites, and quotes and purchases are where agencies balk at self-service.
Who holds the license when software sells
Model laws from the National Association of Insurance Commissioners (NAIC) bind no one until a state enacts its own version, which can differ. NAIC’s Producer Licensing Model Act bars an individual or a business entity from selling, soliciting or negotiating insurance without a license, and has no category for software. An agency’s chatbot therefore sells under the agency’s license, if what it does counts as selling or soliciting.8
The licensing model does not require an insurance company itself to hold a producer license, and exempts employees paid no commission whose function is underwriting or claims. Lemonade’s insurance agency subsidiaries hold licenses to sell, and the company says the employees who negotiate coverage terms are mostly underwriters, who need no agent license.5 So the licensing model will not by itself keep a licensed agent in each sale.
No regulator we found has said whether a chatbot’s quote counts as soliciting. In a 2006 opinion, New York’s insurance department held that call center staff quoting auto premiums needed agent licenses, since quoting meant interviewing prospects and discussing terms.9 We think that reasoning could reach a chatbot doing both. Texas’s June 2026 AI bulletin, addressed to regulated entities and their agents, expects a person to review and agree with any consequential AI decision first, without defining consequential.10
When an agency’s own AI meets the rules
A lead-scoring model, quoting tool or chatbot treats every lead the way it was set up to. That matters under NAIC’s model Unfair Trade Practices Act, which applies to any person engaged in the business of insurance, licensed or not, and for its own purposes counts producers as insurers. A practice it lists becomes a violation only if committed flagrantly and in conscious disregard of the law, or often enough to indicate a general business practice.11 On our reading, a tool set up to carry out a listed practice would meet that second test by design.
The trade practices model’s unfair discrimination section concerns the policy itself: rates, premiums, benefits and terms, and refusing or limiting coverage on listed grounds such as race, religion or national origin. Only its health insurance item adds discrimination “in any other manner.” Marketing comes in elsewhere: insurers and producers may offer value-added services, and gifts within a cost limit, when marketing or selling insurance. Among the conditions, the offer must not be unfairly discriminatory, and any value-added service must rest on documented objective criteria kept for the department to inspect.11
Whether a tool discriminates unfairly is for state law and the facts. We would check whether it picks who is offered a gift or value-added service, whether it feeds a rate, term or coverage decision, and whether it relies on a listed ground. For life and health policies, avoiding listed grounds is not enough: the section also bars unfair discrimination between people of the same class and risk. Read literally, neither provision reaches a tool that only decides which prospects hear about which product, with no offer attached, except in health insurance.
States have other discrimination laws, and California’s insurance department addressed that use in a 2022 bulletin on racial bias and unfair discrimination. It told insurance companies and other license holders to avoid bias that can result from AI in marketing, warned that relying on algorithms to decide whom to market to risks denying eligibility on protected grounds, and required due diligence before using any marketing tool.12
NAIC’s AI bulletin, which state insurance departments can issue as their own, is written for insurance companies alone.13
South Carolina’s 2026 bulletin is addressed to producers and other license holders as well as insurance companies, and holds companies that use AI responsible for what their producers decide and do based on its output. North Dakota’s puts the duty on insurance companies, whose written programs for using AI are to cover AI their authorized agents use on their behalf for decisions affecting consumers. Agencies there meet it through how their carriers govern their distribution channels.14
Whatever share of an agent’s work AI takes on, the agency using it still sells under its own license and answers for what its tools do. So it checks its tools first, then finds its state in our state tracker and uses the self-check on whether rules written for insurance companies reach it. It also asks each carrier it represents what that carrier’s written program for using AI expects of agents.
Footnotes
-
US Bureau of Labor Statistics, “Employment Projections: 2025-2035 Summary,” news release, August 27, 2026, https://www.bls.gov/news.release/ecopro.nr0.htm ; “AI exposure categories,” page and workbook, last modified August 27, 2026, https://www.bls.gov/emp/publications/ai-exposure-categories.htm . The workbook rates insurance sales agents (41-3021) “Very high,” the highest of four relative groups. ↩ ↩2 ↩3
-
US Bureau of Labor Statistics, Occupational Outlook Handbook, “Insurance Sales Agents,” last modified August 27, 2026, Job Outlook tab and projections table, https://www.bls.gov/ooh/sales/insurance-sales-agents.htm . The table runs from 572,600 jobs in 2025 to 591,400 in 2035. ↩ ↩2 ↩3 ↩4
-
Christine Machovec, Michael J. Rieley and Emily Rolen, “Incorporating AI impacts in BLS employment projections: occupational case studies,” Monthly Labor Review, February 2025, https://www.bls.gov/opub/mlr/2025/article/incorporating-ai-impacts-in-bls-employment-projections.htm . The article’s case studies use the 2023-33 projections, which built in AI-related effects for several occupations. ↩
-
Carl Benedikt Frey and Michael A. Osborne, “The Future of Employment: How Susceptible Are Jobs to Computerisation?” Oxford Martin School, September 17, 2013, https://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Future_of_Employment.pdf . Insurance sales agents are number 565 of 702 in the appendix ranking (p. 69); the authors set the high-risk group at probabilities above 0.7 (pp. 37 to 38). ↩ ↩2 ↩3
-
Lemonade, Inc., Form 10-K for fiscal year 2025, filed February 25, 2026, Item 1 and the “Required Licensing” and “Licensing of Our Employees and Adjusters” passages under Insurance Regulation, https://www.sec.gov/Archives/edgar/data/1691421/000169142126000016/lmnd-20251231.htm . The same licensing passage lists two insurance company subsidiaries among the entities that must hold licenses. ↩ ↩2 ↩3
-
US Bureau of Labor Statistics, Occupational Outlook Handbook, “Insurance Underwriters,” https://www.bls.gov/ooh/business-and-financial/insurance-underwriters.htm ; “Claims Adjusters, Appraisers, Examiners, and Investigators,” https://www.bls.gov/ooh/business-and-financial/claims-adjusters-appraisers-examiners-and-investigators.htm ; “Financial Clerks,” https://www.bls.gov/ooh/office-and-administrative-support/financial-clerks.htm ; each last modified August 27, 2026. Projected change from 2025 to 2035: underwriters down 4 percent; claims adjusters, examiners and investigators down 5 percent, and down 6 percent for the whole claims group, which includes auto damage appraisers, whose AI exposure rating is “High”; insurance claims and policy processing clerks down 2 percent. The underwriters page cites automated underwriting software, the claims page technology that automates some tasks and growing AI capabilities, and the clerks page productivity-enhancing technology. The claims page also points to natural disasters as a source of demand. ↩
-
Independent Insurance Agents & Brokers of America, “Big ‘I’ and Future One Release 2026 Agency Universe Study,” September 23, 2026, https://www.independentagent.com/news/big-i-and-future-one-release-2026-agency-universe-study/ . Conducted by Zeldis Research with Future One, a coalition of the association and carriers including GEICO, Liberty Mutual and Progressive; 1,376 respondents. The association has not published the study’s full method. The same release says most agents still prefer that customer service go through the agency. ↩
-
NAIC, Producer Licensing Model Act (#218), January 2005 edition, Sections 2 (definitions), 3 (license required) and 4 (exceptions), https://content.naic.org/sites/default/files/model-law-218.pdf . Section 4A says nothing in the act requires an insurer to obtain a producer license, and for that purpose “insurer” does not include the insurer’s officers, directors, employees, subsidiaries or affiliates. Section 4B(1)(b) exempts officers, directors and employees paid no commission whose function relates to underwriting, loss control, inspection or the processing, adjusting, investigating or settling of claims. Section 6B, which NAIC marks optional, has a business entity acting as a producer obtain its own license and designate a licensed producer responsible for its compliance. ↩
-
New York State Insurance Department, Office of General Counsel, Opinion No. 06-03-01, March 2, 2006, now on the Department of Financial Services website, https://www.dfs.ny.gov/insurance/ogco2006/rg060301.htm . It answers the inquirer’s specific proposal, an authorized insurer’s call center located abroad, whose staff would quote motor vehicle insurance premiums to New York callers, and says that unlicensed individuals or entities who quote premiums “act in the capacity of an insurance agent and must be licensed.” It speaks of individuals and entities, not software. ↩
-
Texas Department of Insurance, Commissioner’s Bulletin B-0003-26, “Use of artificial intelligence,” June 12, 2026, https://www.tdi.texas.gov/bulletins/2026/b-0003-26.html . The bulletin says its goal is not to prescribe specific practices or documentation requirements. ↩
-
NAIC, Unfair Trade Practices Act (#880), Spring 2024 edition, Sections 2 (definitions, including “insurer”), 3 (scope and conditions), 4H (unfair discrimination) and 4I(2) (value-added services and gifts), https://content.naic.org/sites/default/files/model-law-880.pdf . For gifts to individual customers the cost limit is an amount the commissioner finds reasonable; for gifts to commercial or institutional customers the cost must be reasonable in comparison to the premium or proposed premium. ↩ ↩2
-
California Department of Insurance, Bulletin 2022-5, June 30, 2022, https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf . The bulletin names marketing tools alongside data collection methods, fraud algorithms and rating or underwriting tools. ↩
-
NAIC, “Model Bulletin: Use of Artificial Intelligence Systems by Insurers,” adopted December 4, 2023, https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf ; adoption map, https://content.naic.org/sites/default/files/legal-adoption-map-ai-model-bulletin.pdf . Status as of October 8, 2026: the map counts 29 adopting jurisdictions, 27 states plus Puerto Rico and the District of Columbia. It lists South Carolina’s and North Dakota’s 2026 bulletins among the adoptions, and California (the bulletin in note 12), Colorado, New York and Texas as having insurance-specific regulation or guidance of their own; Texas’s entry is a 2020 bulletin, not the 2026 one in note 10. The map does not judge whether each state’s text contains every element of NAIC’s bulletin. ↩
-
South Carolina Department of Insurance, Bulletin 2026-07, “Use of Artificial Intelligence Systems in Insurance,” September 24, 2026, https://doi.sc.gov/DocumentCenter/View/15483/Bulletin-Number-2026-07-Use-of-Artificial-Intelligence-Systems-in-Insurance ; it calls all its addressees “Insurers,” and the responsibility it places on insurers that use AI extends to decisions and actions by their producers, adjusters and other authorized representatives. North Dakota Insurance & Securities Department, Bulletin 2026-2, October 5, 2026, definitions and section 1.7, https://www.insurance.nd.gov/sites/default/files/documents/Bulletins/2026/Bulletin%202026-2%20-%20Use%20of%20Algorithms%2C%20Predictive%20Models%2C%20and%20Artificial%20Intelligence%20in%20Insurance.pdf . ↩
The Bottom Line
- BLS projects agent jobs to grow about 3 percent from 2025 to 2035, assuming technology keeps changing at its past pace.
- BLS also puts agents in its top AI exposure group, meaning AI could do or assist more of their tasks than in most occupations.
- BLS expects underwriter, claims adjuster and claims clerk jobs to shrink, partly through technology, while agents will still be needed to help clients choose.
- NAIC's model for state unfair trade practices laws covers agencies too, and an AI tool applied to every lead can repeat an unfair practice often enough to make it a violation.
- NAIC's AI bulletin speaks only to insurance companies; some 2026 state bulletins reach agents, directly or through their carriers' duties.
Simon Li · Founding Editor
I write InsureAI Wire and maintain its 51-jurisdiction tracker. Most of the work is reading: NAIC working group papers, state bulletins, bills, court filings, and public comment letters. Every claim on the site carries the document it came from, so you never have to take my word for it.
Free · Weekly
Track these developments weekly
Every state and NAIC move on insurance AI, with the source linked, in one email a week.
Free weekly · No spam · Unsubscribe anytime
Information aggregation and analysis, not legal advice. See our disclaimer.