AI can quote insurance — but it still can't replace a good agent

Everybody thought AI was going to take over the insurance agent. The industry has been predicting it for years. More and more sales were moving online, chatbots were getting smarter, and the whole business model felt like it was about to be disrupted.

Here's what actually happened: online insurance sales have flatlined. Not increased, not kept growing — leveled off. AI is speeding up the online transactions that already existed, but it's not pulling new customers out of the agent model. And now there's a growing study showing that the agents who become advisors — the ones who answer the complicated questions AI can't — are going to come out of this transition in better shape, not worse.

AI is good at simple. It's terrible at complicated.

Here's the fundamental problem with AI for insurance: insurance is almost never a simple calculation. Our favorite line on this show is "it depends" — because it almost always depends. What's covered? Depends. What should your deductible be? Depends. What limits do you need? Depends. On your assets, your family, your risk tolerance, your state, your industry, what you're actually trying to protect against.

AI can pull from an FAQ page. It can tell you what your deductible is on an existing policy. It can send you a basic quote based on a few inputs. What it can't do is sit across from you and ask the follow-up question nobody told you to ask.

The chatbot loop everyone hates

You know this experience. You call your insurance company (or any company, really), and the first thing you get is a chatbot asking what you're calling about. You type in your issue. The chatbot responds with a link to an FAQ. "Did this help?" No. It did not help.

You type "representative." Then you type it again. And again. You finally get transferred to a human — who asks you to repeat the same information you already typed into the bot. The bot didn't pass it along. The human asks what your policy number is. You have to find it again.

This is supposed to be an improvement over the old 1-800 phone tree. In some ways it's faster. In the ways that actually matter — getting a complex question answered by someone who can see your whole situation — it's often worse.

The dentist example that explains everything

There's a dentist going viral on social media right now who's calling out an underappreciated part of this: she has to hire full-time staff just to call insurance companies for pre-authorizations. Not because the information is hard to get. Because the carriers refuse to put it in an online portal where doctors could look it up themselves.

So her office makes a phone call. They talk to a human at the carrier who literally pulls up their own internal portal and reads off whether a procedure is approved. The information exists. It's just gatekept behind a phone call that's required to burn 20 minutes of her staff's time, every single time.

This is AI applied to the wrong problem. Instead of using technology to make the partnership between provider and payer better, carriers used it to reduce their internal costs — and shifted the burden onto doctor's offices to hire whole new categories of staff to deal with it.

Why insurance carriers don't actually want to fix this

Honest take, and it's not just insurance: this is a corporate pattern across most big industries.

CEOs and executives change roles every few years. Each new person takes a job with new bonus structures tied to new metrics. To justify their position, they have to show they changed something. So they implement new initiatives, hit the new numbers, and move on — usually before the long-term impact becomes visible.

Nobody's bonus is tied to "maintained a healthy, long-term relationship with providers and customers." Everyone's bonus is tied to short-term metrics that look good in a quarterly report. So the cheese keeps moving, service keeps getting worse, and the people on the ground — agents, doctors' offices, customers — bear the cost.

The role AI should play

We're not anti-AI. There are places where it's genuinely useful:

• Processing routine transactions: Getting an ID card, paying a bill, updating a mailing address.
• Pulling policy documents fast. If you need your declarations page right now, a well-designed chatbot can hand it to you in 30 seconds.
• Answering extremely common questions. "What's my deductible?" "When is my next payment due?"
• Freeing up human agents to work on complex cases. If AI handles the simple stuff, the humans can spend more time on the questions that actually need judgment.

Where it fails: anything involving a judgment call. Any coverage question with "it depends" in the answer. Any situation where the client doesn't know what they don't know and needs someone to ask the right follow-up.

The agent-as-advisor model

The insurance agents who are going to thrive in the next decade are the ones who stopped thinking of themselves as salespeople and started thinking of themselves as advisors. People who ask about your business, your family, your assets, your worries — and build a coverage plan around all of it, not just the one policy you called about.

That's not something AI replicates. Not because AI isn't powerful enough, but because the whole exchange depends on trust, context, and the kind of memory that comes from working with the same person for years. A chatbot that doesn't know your kid just got their license, or that you added a second business location, or that you've been thinking about adding an umbrella because you joined a nonprofit board — can't give you the right advice.

The honest bottom line

AI is here, it's going to keep improving, and it's going to keep handling more of the routine work. That's fine. The question for consumers isn't "AI or human agent" — it's "am I actually getting good advice, and would I know if I weren't?"

If your insurance situation is simple and you're genuinely well-covered, AI-driven direct carriers might be perfectly fine for you. If your life is at all complicated — rental properties, teenage drivers, a business, assets you'd hate to lose in a lawsuit — you want a human who gives a damn and who's going to ask the question nobody else is asking.