Open a browser, search a question about health plan funding, and an AI summary answers before you ever reach a link. That's often the starting point for every renewal conversation now, so the useful question isn't whether to use it. The useful question is how to use it well.
There's no shame in bringing an AI conversation to your broker. Say so plainly. "Here's what I asked Claude, here's what it told me, and I don't understand half of it" is a strong opening. It gives your broker something concrete to respond to, and it saves you from pretending to be an expert on a subject you may or may not actually know well.
Some brokers get defensive about this. Consider that a signal. A broker who can't engage with an outside perspective on your plan probably isn't engaging with much else either. The better response is to treat the AI output as a third voice in the room. Experts are experts in what they do, and even a good one doesn't see everything.
The most common failure isn't a wrong number. It's a right number applied without context.
A few years ago, a wave of employers arrived at renewal funding exactly 50 percent of the employee-only rate, straight across the board, with no strategy behind it. Trace that back and you find technical writers who turned insurance regulations into blog posts, and AI tools that read those posts as instruction. The underwriting manual became the strategy. An underwriting manual is a minimum, not a plan.
Funding at the minimum can be the right call when affordability is tight. What it doesn't do is leave room for strategy. If your renewals have historically climbed, you need budget room for next year built into this year's decision, plus a view on what cost containment might bend the curve. A tool answering only the question you asked won't raise that on its own.
AI output is about as good as the prompt behind it, and most of us overestimate how clear our prompts are. There's no facial expression, no pause, no shared history of your company. Everything it needs, you have to say.
The fix is repetition, and this is where AI beats delegating to a person. Asking a colleague to redo work you briefed badly feels terrible. Asking a tool to redo it four times while you sharpen the question costs nothing. Ask it to ask you clarifying questions before it answers. Ask what sources it used, then go check them.
Fill in the brackets and paste these into whichever tool you use. The first one is the long one. Start there.
1. Brief it on your situation.
You are helping me prepare for our employee health plan renewal. Here is our situation:
- Company: [industry], [number] employees, [state or metro]
- Funding: [fully insured, level funded, or self funded]
- Current design: [plan types, deductibles, employer contribution strategy]
- Renewal history: [last three renewal increases]
- Claims context: [loss ratio, known high cost claim or specialty drug exposure, de-identified]
- Business context: [growth plans, any merger or acquisition in the next 12 to 24 months, ownership or generational transition]
Ask me clarifying questions before you answer. Then give me:
1. The increase we should reasonably expect given our history, industry, and region, and what you are basing that on.
2. Three funding or plan design strategies worth discussing, including one we have probably not considered.
3. The pitfalls of each.
4. The specific questions I should ask my broker, our PBM, and our TPA.
5. Your sources for anything you state as a benchmark.
2. Pressure-test what your broker brings back.
Here is the renewal my broker presented: [rates, plan design changes, employer contribution]. Here is what we had last year: [same detail].
Tell me what actually changed and what that implies. Then give me the three questions a benefits consultant would push back on, and the information I am still missing to evaluate this properly.
3. Hold your downstream vendors accountable.
We are [self funded or level funded] with [TPA] administering claims and [PBM] managing pharmacy. Our biggest cost drivers are [categories].
List the specific questions I should ask each vendor this quarter. For each one, tell me what a good answer sounds like and what a deflection sounds like.
You don't control your TPA or your PBM directly, but you pay them, and AI is now good enough at self-funded mechanics to tell you where to push.
Rates and plan designs are fine. High claimant reports, incurred but not reported billing, and named prescription drug reports are not. De-identify before anything goes in, and check for hidden columns on the spreadsheet you're about to upload. It's easy to hand over more than you meant to.
The strongest use of AI here isn't replacing your broker. It's arriving with better questions, and getting to the strategic conversation faster than the two of you would have on your own. Some of what it hands you will be noise. Some of it will be the discussion nobody had time for last year.
Hear Bret Brummitt walk through how he uses AI with clients on the Generous Benefits Podcast: https://generousbenefits.podbean.com/e/ask-better-questions-use-ai-to-supercharge-benefits-conversations/