AI for Insurance Agents · Extended Session

The 8% Club: How Insurance Agents Actually Put AI to Work

A field report, not a sales pitch. Nobody sponsored it, and there's no software to buy at the end.

By Bill Murray, Wisconsin licensed insurance agent since 1995 · Published August 3, 2026

Ninety-Eight and Eight

Ninety-eight percent of insurance agencies say they're investing in AI this year. Eight percent actually use it every day.

Ninety-eight… and eight. That gap is the most interesting number in our industry right now, and it's the reason for this session. Not the ninety-eight — the eight. Who they are, what they actually do all day, and how your agency slips into the club without hiring a consultant or buying a single robot.

I've run CEWisconsin, an online CE school for Wisconsin insurance agents, since 2002. For the last couple of years I've been running a small AI lab out of it. Some wins, a few faceplants, all of it useful. So take this as a field report rather than a sales pitch — nobody sponsored it, and there's nothing to buy at the end.

Everybody's Standing in the Shallow End

The Big "I" Agents Council for Technology surveyed agencies this year. About two-thirds plan to increase their AI use over the next twelve months. But when you ask what agencies are doing right now, the picture changes: 31% aren't using AI at all, roughly another third describe themselves as "just experimenting," and only about 8% have it genuinely baked into the daily routine.

Vertafore asked a different question of more than 1,300 agency professionals — not what they're doing, but how they feel about it. Thirty-nine percent describe themselves as curious and exploring use cases, 29% as cautious and waiting for proven results, and 9% as outright skeptical.

So if you feel behind: breathe. Almost everybody is standing in the shallow end together, telling each other how great the water is.

The Question Changed This Year

Two years ago, the question at every roundtable was "what is this stuff?" In 2026 the question is "what does this actually do to my workflow — and where does it put my E&O exposure?"

That's a better question. That's a grown-up question. It means we're past the magic show and into the chore list. And the chore list is exactly where the 8% Club lives.

If a salesperson can't tell you which chore it does, that's a bingo, not a solution.

Two honest translations before we go further, because the vocabulary is doing a lot of work in vendor demos. An LLM is the engine inside tools like ChatGPT — think of it as a very well-read intern who has never met your clients. "Agentic" means software that takes several steps on its own — an intern you've let touch the filing cabinet. Most of the rest of the vocabulary is seasoning.

What the 8% Are Not Doing

Start with the negative space, because it's more useful than the highlight reel.

• They did not put a robot on the phones.
• They did not spend six figures on a custom build.
• They did not replace anybody — most are trying to hire, same as you.
• They did not put a consultant on retainer.

The membership dues are embarrassingly low. What they built is smaller and duller than the LinkedIn version, and that is precisely why it works.

What the 8% Are Doing

The boring win	What it looks like in practice
Policy checking	Comparing the renewal against the expiring and flagging what moved — limits, forms, endorsements. The agent still reads it; AI just says where to look first.
Summarizing calls and notes	Turning scribbled notes into a clean, structured file entry, so the file actually tells the story six months later.
Drafting client emails	Renewal reminders, coverage-change explanations, non-renewal notices. AI writes the scaffolding; a licensed human fixes it and sends it.
Sorting the inbox	Triaging what's urgent, what's a service request, and what can wait until Thursday.
Reconciling commission statements	The monthly chore nobody volunteers for, done in a fraction of the time.

Notice the pattern. These are the chores that eat your afternoons. The 8% aren't doing anything flashy — they're doing the same chores you do, faster. It's a supper club, not a spaceship.

My Receipts

So you know I'm not preaching from a brochure: there's an AI on this website. His name is CEWiz. He answers agents' questions about courses, credits, and deadlines around the clock — including two in the morning, which is apparently when Wisconsin agents remember their CE is due. No judgment. I've been there. Every question he answers is one I don't type myself, and that adds up to real hours back in my week.

And here's the part that keeps him honest — because he wasn't honest on day one. Early on I caught him giving an agent a wrong answer. Confidently. Politely. Wrong. I found it because I read the transcripts, which I still do, and I rewrote his training file that same night.

That's the pattern for any AI you let near your clients: short leash, your documents, your review, and a human within arm's reach.

Four Ways This Can Work Against You

The Rulebook Has Arrived

This is no longer a purely voluntary conversation. The NAIC issued a model bulletin on insurers' use of AI systems, and on March 18, 2025, Wisconsin's Office of the Commissioner of Insurance adopted it — making Wisconsin the 24th state to do so.

And yet: more than half of agencies have no written AI policy at all. Not a binder. Not a page. Nothing.

Thirteen percent. Fewer agencies have a written AI policy than have AI in daily use — which means some of the 8% are running AI every day with nothing governing it.

The One-Page AI Policy

Here's the cheapest E&O protection you'll buy this year, and it's free. Four lines on one page.

1. Approved tools. A short, named list. "ChatGPT, Claude, and Gemini are approved for the uses below. Nothing else without approval." Vagueness here is what produces shadow AI.
2. What never goes in. Client personal information, health details, policy and account numbers, Social Security numbers. The postcard rule applies: if it identifies a real person, it doesn't go in the box.
3. AI drafts, licensed humans decide. Every time, no exceptions. This single line is the difference between an assistant and a liability.
4. Save your work. Keep the draft and the final, so the file shows a human made the call. This is the line that does the work in a deposition.

One page. Write it this week. Your future self — and possibly your future attorney — says thanks.

What AI Still Can't Do

Some perspective, because the tool has limits your license doesn't.

AI can't smell a flooded basement. It can't stand in a driveway and notice the boat behind the garage that never made it onto the application. It doesn't know that Earl's "shed" has a full bar and a big screen inside — you know that, because Earl invited you over. And it cannot sit across a kitchen table from a widow with a file folder and know when to stop talking.

Your judgment is the product. AI is the intern. A good intern. Still an intern.

Three Moves This Month

1. Pick one boring chore. Summarizing call notes is a good starter. Use AI on it for two weeks — one chore, not ten.
2. Write the one-page policy. You've seen all four lines. It's a coffee-break job.
3. Build the habit. Nothing AI-drafted leaves the agency unread by a licensed human, and the file keeps the evidence.

Do those three before renewal season gets loud, and you're doing more than ninety percent of this industry.

The 8% aren't smarter than you — and frankly, they aren't better looking either. They just started smaller. One chore. One page. One habit.

Frequently Asked Questions

What percentage of insurance agencies use AI every day?

About 8%. The 2026 Big "I" Agents Council for Technology (ACT) Tech Trends Report, released February 2026, found only 8% of independent agencies have AI embedded in their daily workflows. The same survey found 31% are not using AI at all and roughly a third describe themselves as "just experimenting," while about two-thirds plan to increase AI use over the next twelve months. Separate Liberty Mutual data reported by Risk & Insurance also puts daily AI use among independent agents at 8%.

Does Wisconsin regulate how insurance agents use AI?

Wisconsin's Office of the Commissioner of Insurance issued a bulletin titled "The Use of Artificial Intelligence Systems in Insurance" on March 18, 2025, making Wisconsin the 24th state to adopt the NAIC Model Bulletin on insurers' use of AI systems. The bulletin is addressed to "All Insurers and Interested Parties," and its written AI Systems Program requirement applies to insurers rather than to independent agencies. However, its opening reminder is directed at all entities regulated by OCI — and agents are regulated by OCI. There is no rule requiring an independent agency to have a written AI policy, but when both your regulator and your carriers have published written AI expectations, an agency operating with nothing in writing is the weakest link in the chain.

Does an insurance agency need a written AI use policy?

No Wisconsin rule requires one, but more than half of agencies have nothing in writing at all. The 2026 ACT Tech Trends Report found 55% of agencies have no written AI use policy, 23% have one in development, and only 13% have adopted a formal policy. A written policy is the document that shows a carrier, a regulator, or an E&O defense attorney that AI use in your agency was governed rather than improvised. It fits on one page and needs only four lines: which tools are approved, what information never goes in, that AI drafts while a licensed human decides, and that you keep both the draft and the final.

Can using AI create an E&O claim for an insurance agent?

Yes, in two main ways. First, unverified output: AI states incorrect information with complete confidence, and if an agent passes a wrong coverage answer to a client without checking it, the agent owns that error — not the software. Second, shadow AI: Gartner research reported by Risk & Insurance found 88% of employees with enterprise AI access also use personal AI tools, meaning client information can flow through accounts the agency cannot see, log, or produce later. If AI helped write advice and the advice was wrong, "the computer did it" is not a defense.

What is the easiest way for an insurance agent to start using AI?

Pick one boring chore you already do every week, that contains no client personally identifiable information, and that you can check in under a minute. Summarizing call notes is a good starter. Use a free browser-based tool such as ChatGPT, Claude, or Gemini, and do that one chore for two weeks before adding a second. Agencies with AI in daily use didn't start bigger than this. They started smaller and kept going.

How do most consumers feel about insurance companies using AI?

Cautiously. Pew Research Center found that 50% of U.S. adults say the increased use of AI in daily life makes them feel more concerned than excited, compared with just 10% who are more excited than concerned and 38% who feel equally both. That matters for agencies because many clients call precisely because they want a human being. Using AI behind the scenes to work faster is sensible; using it to put distance between you and the client works against the reason they called.

Back to the Series Back to Blog