> An automated insurance agent is AI that runs the sales conversation for a licensed agency or carrier. What it can sell, step by step, and the rules.

Canonical: https://www.sellence.com/guides/ai-for-insurance-agencies

Title: Automated Insurance Agent: What AI Can Sell, Step by Step

Insurance Guide

# AI for Insurance Agencies and Carriers: What an Automated Insurance Agent Can Sell, Step by Step

By [Guy Frum](https://www.linkedin.com/in/guy-frum), Head of Growth Updated 11 October 2026 14 min read

Short Answer

An automated insurance agent is AI software that runs the sales conversation with a prospect or policyholder: the first reply to a quote request, the quote, the follow-up, the objection and the buying decision, by text, chat or voice. It works for a licensed agency or carrier, under that license and its supervision, and state rules decide which steps the licensed producer of record must own.

Captions are on screen. Tap Sound On to hear it, or read the transcript below.

**Read the Transcript**

Most agencies use AI to save time. Only 17% expect it to grow revenue. The revenue is in the sale itself. An automated insurance agent answers a new lead in seconds. Sends the quote. Follows up. Handles the objection. Gets the decision. It works under your agency's license, and your state sets which steps a licensed agent owns. See what it can sell, step by step. Read the guide on sellence.com.

It already exists. Lemonade's annual report for 2025 says "AI Maya and our APIs sell 98% of Lemonade's policies" ( [Form 10-K](https://www.sec.gov/Archives/edgar/data/1691421/000169142126000016/lmnd-20251231.htm), filed 2026-02-25). Those sales run through Lemonade's licensed agency subsidiaries, and 382 of its employees held insurance licenses at the end of 2025.

Most search results for the term describe AI for claims, underwriting and back-office work. This guide covers the sales side: what saves time vs what sells, each step of a sale, an example conversation, and the license, AI and consent rules.

## Saves Time vs Sells: Where Agencies Use AI Today

Most agency AI today saves staff time. Very little of it talks to buyers.

The 2026 Agency Universe Study from the Big I and Future One, conducted by Zeldis Research with 1,376 respondents, found 46% of independent agencies use AI, up from 15% in 2024 ( [Insurance Journal](https://www.insurancejournal.com/news/national/2026/09/24/886639.htm), 2026-09-24). The top uses: marketing content (49%), coverage form analysis (43%) and contract review (35%).

The Big I's Agents Council for Technology asked agencies what they expect AI to deliver ( [2026 Tech Trends Report](https://www.independentagent.com/news/two-thirds-of-independent-agents-plan-to-increase-ai-use-this-year/), released 2026-02-19). Operational efficiency came first at 60%, staff productivity 52%, customer experience 21% and revenue growth 17%. In the same survey, 8% had AI embedded in daily workflows, 13% used AI chatbots or virtual assistants, and 55% had no written AI use policy.

The February report found 31% of agencies not using AI and the September study found 46% using it. Different samples and questions, so read each on its own terms.

| Saves time (agent-facing) | Sells (prospect-facing) |
| --- | --- |
| Marketing content: 49% of AI-using agencies (Agency Universe Study 2026) | First reply to a new quote request, at any hour |
| Coverage form analysis 43%, contract review 35% (Agency Universe Study 2026) | Quote follow-up over several days |
| Public models such as ChatGPT: 45% of agencies (Big I ACT 2026) | Objection handling: price, timing, comparing options |
| Document and data extraction: 13% of agencies (Big I ACT 2026) | Renewal re-shop and retention conversations |
| Claims intake: Lemonade's AI Jim takes 96% of first notices of loss without a human (10-K) | End-to-end digital sale: Lemonade's AI Maya and APIs sell 98% of its policies (10-K) |

Buyers already use AI to shop. In J.D. Power's first U.S. AI Insurance Experience Study, fielded June to July 2026, 29% of auto and home customers used AI tools, and 42% of customers who used AI to shop for a new policy bought one ( [Insurance Journal](https://www.insurancejournal.com/news/national/2026/08/28/883247.htm), 2026-08-28). A Big I consumer survey of 400 insured adults, run by MFour, found 87% say a human agent remains important and 61% are more likely to choose an agent who uses AI ( [Big I](https://www.independentagent.com/news/keeping-a-human-in-the-loop-remains-crucial-to-insurance-buyers-as-they-accept-ai-tools/), 2026-09-09).

Evidence that the right column sells is thin. Lemonade's 98% is one direct carrier. McKinsey describes one unnamed carrier whose after-hours chatbot produced an 11% increase in prospects who bought policies ( [McKinsey](https://www.mckinsey.com/industries/financial-services/our-insights/the-future-of-ai-in-the-insurance-industry), 2025-07-15). No public study measures AI-led quote follow-up, objection handling or renewal conversations against bound premium at US agencies. Until one exists, your own controlled test is the benchmark.

## What an Automated Insurance Agent Can Sell, Step by Step

This guide splits a sale into 6 steps. For each: what the model does, where the license line sits in states with published guidance on unlicensed staff, and the number to watch.

Where the license line sits varies by state. This follows the New York, Florida and Oklahoma guidance in the table below.

### 1. Lead In

A quote request comes in from the website, a comparison site or a renewal list. The model replies in seconds, says it is AI, collects the facts the quote needs and books a call if the buyer wants one. New York, Florida and Oklahoma guidance treat taking information and scheduling as clerical work.

Watch: minutes from request to first reply, and the share of requests that get a 2-way conversation.

### 2. Quote

The model sends the rate the agency's or carrier's rating system produced. Oklahoma treats quoting as clerical when the rate comes from a published or computer database of standard rates, on request, with no advice on coverages, deductibles or limits. Lemonade's Maya goes further and creates the quote itself, inside a licensed structure.

Watch: quotes delivered, and minutes from request to quote.

### 3. Follow-Up

The model follows up across days, picks up from the last message, switches between text and call inside the consent the buyer gave, and answers factual questions from the quote. The [quote-to-bind guide](https://www.sellence.com/guides/quote-to-bind) maps where quotes drop off before bind.

Watch: follow-up rate, quoted buyers reached divided by quotes issued.

### 4. Objection

Price, timing, "I need to talk to my spouse", "I'm comparing". The model answers with facts from the quote, prices the changes the buyer asks for, sends a side-by-side to forward and sets a time to talk again. Discussing coverage, limits and deductibles is licensed activity in New York, Florida and Oklahoma guidance, so the model works this step under the agency's license, and the agency's compliance owner sets how far it goes per state.

Watch: the objection reason on every unbound quote, and bind rate after follow-up.

### 5. Decision

The buyer picks a version and a start date. Recommending a policy is licensed: Oklahoma's guidance says that if a prospect asks for advice, "the response must be made by a licensed producer". So a question like "which one would you pick?" goes to the producer of record in states that draw the line there.

Watch: decisions per quoted buyer reached, and days from quote to decision.

### 6. Bind

The model sends the application and payment link with the buyer's start date. Binding happens under the agency's or carrier's license. Where the state requires a licensed individual to bind, as Oklahoma does for signing binders, that step belongs to the producer of record.

Watch: quote-to-bind ratio, and the drop between the payment page and bound.

## An Example Conversation (Illustrative)

Illustrative only: names, address and prices are invented. Jordan requested a home quote on the agency's website and gave written consent to texts.

> Day 0, 7:52 p.m.
> Agency: Hi Jordan, this is Sam, the AI assistant at [Agency]. I have your home quote request for 14 Oak Lane. To price it I need the year the roof was last replaced. Reply STOP to opt out.
> Jordan: 2019

Step 1, lead in: first reply inside a minute, AI disclosed in the first line, opt-out included.

> Agency: Thanks. Your quote from [Carrier] is $1,840 a year with a $1,000 deductible. The full quote is in your email.
> Jordan: thats more than i pay now. what would it be with a 2500 deductible

Step 2, quote: the rate the agency's rating system produced, sent on request.

> Agency: With a $2,500 deductible it's $1,610 a year, everything else the same. Want both versions side by side?
> Jordan: yes. need to run it by my wife
> Agency: Sending both now so you can forward them. When are you both looking at it?
> Jordan: tomorrow night

Step 4, objection: the buyer named the change and the model priced it. The spouse objection got something to forward and a time.

> Day 1, 7:30 p.m.
> Agency: Hi Jordan, did you and your wife get to compare the 2 versions?
> Jordan: yes we'll do the 2500 one. can it start the 1st
> Agency: Yes. Here's your application and payment link, with the 1st set as the start date.
> Application signed and paid. Policy bound under [Agency]'s license, start date the 1st.

Steps 3, 5 and 6: the follow-up picked up from the last message, the buyer decided, and the bind ran under the agency's license, producer of record named on the file.

## What the License and State Rules Require

Producer licensing is state law and varies by state. This is general information, not legal advice. Check each state where you write business.

### The Producer License

The NAIC Producer Licensing Model Act ( [Model #218](https://content.naic.org/sites/default/files/model-law-218.pdf), January 2005 version) says a person shall not "sell, solicit or negotiate insurance" in a state without a license for that line. "Person" means an individual or a business entity. The model defines the 3 acts:

| Act | Model #218 definition (Section 2) |
| --- | --- |
| Sell | "to exchange a contract of insurance by any means, for money or its equivalent, on behalf of an insurance company" |
| Solicit | "attempting to sell insurance or asking or urging a person to apply for a particular kind of insurance from a particular company" |
| Negotiate | Conferring directly with or offering advice directly to a buyer about the substantive benefits, terms or conditions of a contract |

Each state writes its own version. As of 2026-10-11, no state regulator has published guidance on whether AI software itself sells, solicits or negotiates. The FORC Journal ( [Vol. 36 Ed. 3](https://forc.org/Public/Public/Journals/2026/Articles/Fall/Vol36Ed3Article2.aspx), 2025-11-20) lists the open questions, such as whether an "AI agency" needs a designated responsible licensed producer.

### What State Guidance on Unlicensed Staff Says

The closest guidance covers what unlicensed agency staff may do. New York, Florida and Oklahoma have written it down:

| Activity | New York ( OGC Op. 04-12-03, 2004) | Florida ( 69B-222.060, 1993) | Oklahoma ( Bulletin 2013-09, 2013) |
| --- | --- | --- | --- |
| Schedule appointments | Allowed with no discussion of coverage or cost | Allowed | Clerical |
| Take information, pass the request on | Allowed | Allowed | Clerical |
| Give a rate | On the insurer's "may not" list the OGC reviewed | Not addressed directly | Clerical when from a published rate source, on request, with no advice |
| Advise on coverage, limits, deductibles | Not allowed | Never allowed | Licensed |
| Initiate a sale | Not allowed | Never allowed | Licensed |
| Bind | Not allowed | Never allowed | Licensed |

Of the 3, only Oklahoma names software. Its clerical list includes processing requests "through an automated system developed and maintained under the supervision of an insurer or licensed insurance producer". New York's regulator added that the licensee is accountable for the unlicensed staff working for it.

Our reading, until a regulator says more: the automated agent acts for the licensed agency or carrier that deploys it, under its license and supervision. The state draws the line on which steps the producer of record owns: usually advice, recommending a policy and, where required, binding. Lemonade is the carrier version: licensed agency subsidiaries, licensed staff, and an AI that runs the sale.

### The NAIC AI Model Bulletin

The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers on 2023-12-04. As of the NAIC's map dated 2026-10-08, 27 states plus DC and Puerto Rico have adopted it, the latest North Dakota (2026-10-05) and South Carolina (2026-09-24). California, Colorado, New York and Texas have their own insurance AI rules or guidance ( [NAIC map](https://content.naic.org/sites/default/files/legal-adoption-map-ai-model-bulletin.pdf) ).

The bulletin is addressed to insurers and names "marketing, sales and distribution" among the stages where AI is used. It expects a written AI program, controls scaled to potential consumer harm, notice to consumers that AI is in use, and audit rights over third-party AI vendors. Agencies and vendors meet it through carrier contracts. What an AI says in a sales conversation also falls under unfair trade practice law, the same as anything a producer says (our reading).

### AI Disclosure Laws

State AI disclosure laws are new and narrow, and several states have bills pending. This is general information, not legal advice.

| State | Law | In force | What it requires |
| --- | --- | --- | --- |
| California | BOT Act, Bus. & Prof. Code 17940 to 17943 | 2019-07-01 | Disclosure when automated online software (a "bot" under the law) is used to sell with intent to mislead about its artificial identity |
| California | AB 2905, Pub. Util. Code 2874(a)(3) | 2025-01-01 | Autodialed calls must announce when the message uses an AI-generated voice |
| New Jersey | N.J.S.A. 56:18-1 to 56:18-5 | 2020 | Clear disclosure at the outset when automated software communicates about the sale of merchandise or real estate (application to insurance untested) |
| Utah | SB 226, Utah Code 13-75 | 2025-05-07 | Disclose generative AI in a consumer transaction on a clear request, with a safe harbor for disclosing at the outset |
| Maine | 10 M.R.S. 1500-DD | 2025-09-24 | Clear notice in any trade or commerce where a reasonable consumer could think they are talking to a human |

Colorado is the state to watch. Its chatbot disclosure law (HB26-1263, signed 2026-05-29) starts 2027-01-01 and excludes services built for commerce, shopping, payments and customer service. Its automated decision law (SB26-189, signed 2026-05-14) replaced the 2024 Colorado AI Act before it took effect and covers consequential decisions, insurance included, from 2027-01-01. A sales conversation that makes no underwriting or eligibility decision is likely outside it (our reading).

The practical rule: say it is AI at the start of the conversation and whenever the buyer asks. That meets the Utah safe harbor and the Maine and New Jersey disclosure rules, and removes the intent element in California's BOT Act (our reading).

## Consent for Texts and Calls

Texting and calling a quote request needs consent, captured on the quote form. The [quote-to-bind guide](https://www.sellence.com/guides/quote-to-bind) covers written consent, opt-out words, the 10-business-day opt-out rule and calling hours.

AI voice adds a rule. In FCC Declaratory Ruling 24-17 ( [FCC](https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf), released 2024-02-08), the FCC held that AI-generated voices are "artificial" voices under the TCPA. AI-voice calls need the same prior express consent as prerecorded calls, written consent for telemarketing, plus caller identification and an opt-out for marketing calls. California adds the AI-voice announcement on autodialed calls.

| Channel | Consent | AI disclosure |
| --- | --- | --- |
| Text | TCPA prior express written consent for marketing, opt-out words, calling hours | Maine, Utah on request, New Jersey where it applies |
| Voice | TCPA consent, with AI voice treated as artificial (FCC 24-17) | California announcement on autodialed calls, plus the state rules above |
| Web chat | No TCPA consent needed for an inbound chat | California BOT Act, Maine, Utah, New Jersey |

## How to Evaluate an Automated Insurance Agent

Most "best AI for insurance agents" lists rank productivity tools. For the sales column, score each tool on:

1. Which column it serves: which steps of the sale it runs, from lead in to bind, and which it leaves to your team.
2. The license line: your compliance owner sets, per state, which questions go to the producer of record, named on every file.
3. AI disclosure: it says it is AI in the first message and whenever asked, in your agency's name.
4. Consent and opt-out: it checks consent before every text and call, honors opt-outs on every channel and keeps calling hours in the buyer's time zone.
5. Memory across days: a follow-up on day 3 picks up from what the buyer said on day 1.
6. Objection handling you can read: real transcripts of price, timing and spouse objections, names removed.
7. A control group: bound policies or bound premium against a holdout of the same leads.
8. Carrier requirements: support for carrier AI program reviews, documentation and audit rights in states that adopted the NAIC bulletin.

For tools that convert open quotes, see [AI for insurance quote conversion](https://www.sellence.com/best-ai-insurance-quote-conversion). For policyholder service and claims texting, see [Hi Marley alternatives](https://www.sellence.com/hi-marley-alternatives).

## FAQ

### What is an automated insurance agent?

AI software that runs the sales conversation, from first reply through quote, follow-up, objections and decision, by text, chat or voice. It works under a licensed agency's or carrier's license, and state rules decide which steps the producer of record owns.

### How can insurance agents use AI?

Most agencies use it to save time: marketing content, coverage form analysis and contract review lead the 2026 Agency Universe Study. Fewer use it to sell: replying to quote requests, following up, answering objections and working renewals.

### What is the best AI tool for insurance agents?

Pick by the step of the sale you want covered. Then check the license line in your states, AI disclosure, text and call consent, and whether results are measured against a control group on bound policies.

### Can AI calling agents be used by insurance companies?

Yes, with consent. The FCC ruled in February 2024 that AI-generated voices are artificial voices under the TCPA, so AI-voice calls need the same consent as prerecorded calls, written for telemarketing.

### Will AI replace insurance agents?

The evidence so far points to producers working beside AI. In a 2026 Big I consumer survey, 87% of buyers said a human agent remains important and 61% preferred an agent who uses AI. Licensing also keeps advice and recommendations with a licensed producer.

### Which insurance companies use AI to sell?

Lemonade is the clearest public example: its 2025 annual report says AI Maya and its APIs sell 98% of its policies, through licensed agency subsidiaries. NAIC surveys of large insurers found AI in marketing at 50% of auto insurers (2022) and 47% of home insurers (2023).

Sellence built the persuasion model that runs the sales conversation by text and call under your agency's or carrier's license: it follows up on quotes and renewals for weeks, answers the objection and keeps one thread through the buyer's decision. At Libra Insurance it delivered 3x sales per agent in a controlled 90-day A/B test. See [how it works for insurance](https://www.sellence.com/insurance) or [book a demo](https://www.sellence.com/demo).

## Sources

1. Lemonade, Inc., Form 10-K for fiscal year 2025, filed 2026-02-25. [https://www.sec.gov/Archives/edgar/data/1691421/000169142126000016/lmnd-20251231.htm](https://www.sec.gov/Archives/edgar/data/1691421/000169142126000016/lmnd-20251231.htm)
2. Andrea Wells, "Agency Universe Study" coverage (Big I, Future One, Zeldis Research, 2026 Agency Universe Study), Insurance Journal, 2026-09-24. [https://www.insurancejournal.com/news/national/2026/09/24/886639.htm](https://www.insurancejournal.com/news/national/2026/09/24/886639.htm)
3. Independent Insurance Agents and Brokers of America (Big I), Agents Council for Technology 2026 Tech Trends Report release, 2026-02-19. [https://www.independentagent.com/news/two-thirds-of-independent-agents-plan-to-increase-ai-use-this-year/](https://www.independentagent.com/news/two-thirds-of-independent-agents-plan-to-increase-ai-use-this-year/)
4. J.D. Power 2026 U.S. AI Insurance Experience Study, fielded June to July 2026, reported by Insurance Journal 2026-08-28. [https://www.insurancejournal.com/news/national/2026/08/28/883247.htm](https://www.insurancejournal.com/news/national/2026/08/28/883247.htm) and study page [https://www.jdpower.com/business/u-s-ai-insurance-experience-study/](https://www.jdpower.com/business/u-s-ai-insurance-experience-study/)
5. Big I consumer survey conducted by MFour Data Research, 400 US adults, release 2026-09-09. [https://www.independentagent.com/news/keeping-a-human-in-the-loop-remains-crucial-to-insurance-buyers-as-they-accept-ai-tools/](https://www.independentagent.com/news/keeping-a-human-in-the-loop-remains-crucial-to-insurance-buyers-as-they-accept-ai-tools/)
6. McKinsey and Company, "The future of AI in the insurance industry", 2025-07-15. [https://www.mckinsey.com/industries/financial-services/our-insights/the-future-of-ai-in-the-insurance-industry](https://www.mckinsey.com/industries/financial-services/our-insights/the-future-of-ai-in-the-insurance-industry)
7. NAIC, Producer Licensing Model Act (#218), January 2005 version. [https://content.naic.org/sites/default/files/model-law-218.pdf](https://content.naic.org/sites/default/files/model-law-218.pdf)
8. FORC Journal, Vol. 36 Ed. 3, "Regulatory considerations for the use of AI and technological advancements in insurance transactions", posted 2025-11-20. [https://forc.org/Public/Public/Journals/2026/Articles/Fall/Vol36Ed3Article2.aspx](https://forc.org/Public/Public/Journals/2026/Articles/Fall/Vol36Ed3Article2.aspx)
9. New York Insurance Department, Office of General Counsel Opinion No. 04-12-03, 2004-12-01. [https://www.dfs.ny.gov/insurance/ogco2004/rg041203.htm](https://www.dfs.ny.gov/insurance/ogco2004/rg041203.htm)
10. Florida Administrative Code R. 69B-222.060, Unlawful Activities by Unlicensed Insurance Agency Personnel, adopted 1993-10-12 (text via Cornell LII). [https://www.law.cornell.edu/regulations/florida/Fla-Admin-Code-Ann-R-69B-222-060](https://www.law.cornell.edu/regulations/florida/Fla-Admin-Code-Ann-R-69B-222-060)
11. Oklahoma Insurance Department, Bulletin No. 2013-09, 2013-08-15. [https://www.oid.ok.gov/wp-content/uploads/2019/11/081513_Producer-Licensing-Bulletin-2013-09.pdf](https://www.oid.ok.gov/wp-content/uploads/2019/11/081513_Producer-Licensing-Bulletin-2013-09.pdf)
12. NAIC, Model Bulletin: Use of Artificial Intelligence Systems by Insurers, adopted 2023-12-04. [https://content.naic.org/sites/default/files/cmte-h-big-data-artificial-intelligence-wg-ai-model-bulletin.pdf.pdf](https://content.naic.org/sites/default/files/cmte-h-big-data-artificial-intelligence-wg-ai-model-bulletin.pdf.pdf)
13. NAIC, Implementation of the NAIC Model Bulletin, map with status as of 2026-10-08. [https://content.naic.org/sites/default/files/legal-adoption-map-ai-model-bulletin.pdf](https://content.naic.org/sites/default/files/legal-adoption-map-ai-model-bulletin.pdf)
14. NAIC, Private Passenger Auto AI/ML Survey staff report, 2022-12-08. [https://content.naic.org/sites/default/files/inline-files/PP%20Auto%20Survey%20Team%20Report%20120822.pdf](https://content.naic.org/sites/default/files/inline-files/PP%20Auto%20Survey%20Team%20Report%20120822.pdf)
15. NAIC, Homeowners AI/ML Survey memo, 2023-08-10. [https://content.naic.org/sites/default/files/inline-files/Home%20Survey%20Memo%20to%20BDAIWG.pdf](https://content.naic.org/sites/default/files/inline-files/Home%20Survey%20Memo%20to%20BDAIWG.pdf)
16. California Business and Professions Code 17940 to 17943 (BOT Act, SB 1001), operative 2019-07-01, and Public Utilities Code 2874(a)(3) (AB 2905), effective 2025-01-01.
17. New Jersey, P.L. 2019, c. 486, N.J.S.A. 56:18-1 to 56:18-5, approved 2020-01-21.
18. Utah SB 226 (2025), Utah Code 13-75, effective 2025-05-07. [https://le.utah.gov/Session/2025/bills/enrolled/SB0226.pdf](https://le.utah.gov/Session/2025/bills/enrolled/SB0226.pdf)
19. Maine Revised Statutes, Title 10, chapter 241, section 1500-DD (PL 2025, c. 294), effective 2025-09-24. [https://mainelegislature.org/statutes/10/title10ch241.pdf](https://mainelegislature.org/statutes/10/title10ch241.pdf)
20. Colorado HB26-1263, signed 2026-05-29, and SB26-189, signed 2026-05-14. [https://leg.colorado.gov/bills/sb26-189](https://leg.colorado.gov/bills/sb26-189)
21. Colorado Division of Insurance Regulation 10-1-1 (3 CCR 702-10), amendments effective 2025-10-15, summarized by Faegre Drinker 2025-09-02. [https://www.faegredrinker.com/en/insights/publications/2025/9/colorado-division-of-insurance-expands-ai-related-governance-and-risk-management-obligations-for-insurers](https://www.faegredrinker.com/en/insights/publications/2025/9/colorado-division-of-insurance-expands-ai-related-governance-and-risk-management-obligations-for-insurers)
22. FCC Declaratory Ruling FCC 24-17, CG Docket 23-362, adopted 2024-02-02, released 2024-02-08. [https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf](https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf)

Related

- [Sellence for Insurance](https://www.sellence.com/insurance)
- [Quote-to-Bind Ratio Guide](https://www.sellence.com/guides/quote-to-bind)
- [AI for Insurance Quote Conversion](https://www.sellence.com/best-ai-insurance-quote-conversion)
- [Hi Marley Alternatives](https://www.sellence.com/hi-marley-alternatives)