Using AI to Improve the Conversation with Your Benefits Broker

Welcome to the Generous Benefits Podcast, where we talk about building benefits that actually help people and still make business sense. I'm your host, Amanda Brummitt. Today, we're talking with our own Bret Brummitt about how you can use artificial intelligence to improve the conversation with your benefits broker. Think Claude, Chat, Gemini, Grok, and even Copilot.

Let's jump in. Good morning, Brett Brummett. Good morning. Okay, today, I do want you to introduce yourself to everybody, just in case this is anyone's first time.

But for people that have been listening to the show for a long time, don't give them the same spiel. Today, I want you to tell them, let's give them kind of your techie. You're kind of known as a techie person in insurance. Yeah, so I am the former Employee Benefits News Technology Advisor of the Year of 2018.

So almost a whole, getting close to a decade ago, which now makes me super old news. Irrelevant. Yeah, I know. Aged out of the whole thing.

But have been known a little bit on the techie side. Very mechanical as well. So I really started, you know, started a young age tearing apart my Star Wars toys and putting them back together and seeing how, you know, the X-Wing fighter, all the little lights were wired together in it. Down to my dad's lawnmower equipment when he had a lawn service, tearing those things apart and fixing those.

I actually had and ran a small engine repair service that was my grandfather's retirement gig that he didn't survive long into. So I had the shop and all the tools and the parts. So I turned that into a teenage, preteen, pre-driving hobby that turned into a car hobby that eventually cars got techie enough that turned into a car computer hobby. And at that point in time, I aged back into insurance and have stayed a little bit on the technology front of insurance.

So that's the tech journey. Okay, cool. And then give them the 10-second insurance journey. Wanted to go work for the FBI.

Someone told me to get an insurance license to get tracked. I have been doing insurance ever since then from the old property and casualty days of being an insurance adjuster through selling life insurance to people at their table at house at 7 o'clock at night. Building that super need when one of the spouses has to die so the other one has to survive on money. I do not love doing that.

And quickly learned that I did like doing that to the point that I started hopping into employer offices and getting to meet them and doing health insurance sales. And that was an easy conversation back in 1999. I used to sell insurance plans for $75 an employee, $325 a family. I know all day long I know I'll take three can't even add a zero to those to make it work anymore but been doing it since then got involved in regular health insurance going through all of that the ACA and around the same time of the ACA I transitioned to a lot of self-funded health insurance work ended up with a really large self-funding slide that I got to learn a lot on took me to school and stuff I didn't learn or didn't know at the time thankfully I had a team around me that did and have just stayed on that journey of diving into that and even going getting a master's degree in health plan benefit design around self-funded insurance.

So one of a few people that actually have a weird master's degree in insurance. That's me. Very cool. Very cool.

Okay, well, let's dive into our topic at hand. So I've heard you suggest to clients that they actually use AI to help with your conversations. Outside of the people you've directly said this to, how many clients do you think are using some version of AI to challenge the rates, the plans, whatever you share with them during that renewal process, or even using it to shop rather than directly with their broker? Yeah, I want to say 100% indirect.

I will. I'll say 100% indirectly. And I say that because all of our internet browsers now have a really pretty good, robust AI tool in the area. I mean, if you've gone to any of your searches, you're looking at the AI results before you look at the links these days.

So like that's showing it, showing stuff to us before we even get to the sponsored links, before we get to the regular SEO searches. So indirectly 100 now what i have seen more and more is questions from clients that some are very like copy and paste the screenshot that just doesn't show it in the question some are or does show it or does shot yeah which is fine sometimes it's the weird font spacing back and forth the copy paste like their thoughts ai thoughts in between some edited but you also know it's not exactly their normal tone and pace.

Yeah. But, You know, I've looked at this now, and this has been ongoing probably a good three years, which seems like decades in AI conversation. But it has evolved so well. I mean, I didn't like it at first.

I'm like, why are they doing that? Like, good God, it's just going to be bad advice. Like, now we have to start all over again. Not really the case these days.

And it didn't take me long to go, you know what? They're getting to a point sometimes with some bad advice two or three years ago that got us to a conversation we probably weren't going to unlock anyway yeah and i just said you know what i should embrace this like let's figure this out also let's figure out what ai is learning from like what are those sources that as like a lot of the things originally like i had this this rash of like why Why am I running into all these employer clients, prospects that aren't mine, that are funding at 50% of the employee only rate across the board?

I'm like, okay, well, it looked like, I mean, as I was looking at it, it looked like a lot of the software vendors created in their blogs and their instructions, this, you know, the writers, the technical writers, took all of the insurance regulations and wrote insurance regulations into blog posts that then AI filtered in and said, okay, well, this is how it's done. Not actually how it's done, but it became this race to the bottom of employer funding of the employee cost. And it was happening more and more.

And there was some commonality of which payroll platforms or other platforms those folks were on. But then also AI was given it back when I started testing. I'm like, okay, well, it's not taking the historical what has happened with an employer in your industry in your region or what is the different strategies or what outcome do you want from this funding mechanism? But it was just kind of a copy and paste.

Oh, this is how it's done. Yeah. No, it wasn't how it's done, but that is what the regulation set as a minimum. And, you know, technical writers and AIs weren't looking for context in space, time, history, industry, employee needs.

It was looking at the underwriting manual. And so everything referenced an underwriting manual and underwriting manuals became truth. Not that they're not truth, but an underwriting manual is the minimum. And so everyone is starting at the minimum, which may be the only thing someone can do affordability wise.

However, that's not what the competitors are doing. That's not necessarily the strategy. There's actually sometimes a way to pay less if you don't even structure it that same way. Like there's a whole more robust way of offering a benefit than the underwriting manuals minimum that might not actually cost you as much as doing the minimum would.

So there's more discussion to be had. But again, the AI, to my estimation, my testing was pushing out that underwriting manual minimum, which moved the conversation or moved the experience and honestly kind of put some employers in a little bit of a hurt for funding off a minimum, not having any kind of buffer for what happens in the next two, three, five years with renewal increases. Like, where's the strategy, the budget for, okay, cool, we're here at this rate this year, but, you know, if we just historically look, we got to have a little 7%, 10%, 12% headroom for what's coming down the pike next year.

And so how are we going to get there? Either, you know, benefit reduction cost or budgeting for that, or what kind of cost containment needs to be put in place to kind of buffer it, bend the curve, keep the prices there. Is there something we can do better next year that we're strategically looking for? Do we need to invest more in this one area so we don't have a big issue?

coming down the. So that's where I feel like it was missing a lot at that point. I don't actually feel like it's missing as much today. I feel like the AIs that I've seen and the conversations that I've test run or I've gotten better at the prompts.

Okay, I want to come back to the dangers in just a second. But before I do, I just want to state something as the non-host, non-expert, but there is no shame in bringing an AI conversation to any vendor relationship, other human. To me, where I'm going to laugh at you, and I might even talk bad about you on a podcast, is don't copy and paste something and pretend it's your own knowledge. Like, come at it and say, hey, here's what chat said I should ask you.

I don't understand all this. Explain it to me. You don't need to pretend like you're an expert on it because then all the weird spacing and something maybe not being your language isn't weird if you're like, hey, I don't understand insurance, so I dumped this into AI. Here's what it said.

You know, I will push back a little bit on this. Please do. I have seen lots of brokers being very defensive, not kind of, very defensive of the AI advice, full stop. Okay.

Like, no room. Could it be good? Is this helping the conversation? It's just, you know, you're trusting this and not trusting an expert.

Okay. However, experts are experts in what they do, and they may not see everything. And so if you're adding an AI perspective to it, maybe you're gaining more. Mm-hmm.

Knowledge in a conversation it's like i'm down with it and i actually like it there are things that it does that helps further the conversation that i don't even want to get into sometimes just because i'm like this is not where you want to go you've told me where you want to get to we're going to get there some of it's just noise some of it's actually really good that i may be victim of not spending enough time doing so if you can have a tool that gets me past the conversation that I'm going to either take for granted, gloss over, or maybe don't even want to have just, and it's getting you there and we can have a more robust conversation.

Hallelujah. Yeah. Okay, I'm going to push back on your pushback that if your broker acts like a jerk about it, find a new broker. It is probably going to show up in other areas as well, and I will agree with you on that.

It's probably a red flag about other areas. However, let's leave maybe the experience they've had is the experience of two or three years ago, where the advice was very half-baked and not accurate, or AI got a bad reputation for hallucination, and they haven't had the experience where it's been an additive to the conversation for them. But let's also put in the space that there are definite things in the insurance purchasing stream that it can do, whether it should do, maybe, maybe not.

But it can do. So much like everyone else is worrying about their job and not being relevant anymore, being replaced, not being needed, not being the trusted go-to. There's a little bit of insecurity there, which I think we all deal with, not just, you know, the broker. I appreciate your kind perspective.

Okay, let's talk about the bad stuff, because I eventually want to get to the good stuff. What is the danger in using AI? Is it giving employers bad plan options, bad funding strategies, or their compliance issues with it? I would say that today, and let's, so I'll preface this with, I generally have been hopping, the two main ones I hopped back and forth between are Claude and Gemini.

Okay. Chad is in the realm, Co-Pilot, we've danced, but hasn't been a fun dance for Co-Pilot. My dance hasn't been great either. Yeah.

With steps on my feet a lot. That's like dancing with me. I just choose not to because it's bad. We're cuter than Copilot.

I hope so. A little grok while we're driving in the car, you know, that and forth. But that's my context of it. Where is it bad or where can it go bad?

Yeah. I mean, you got to push on it and ask it to clarify. And this is going to sound bad, but its output is pretty much as good as what you're asking it. And so...

No, no, that's dead accurate. That's not bad. Well, it's bad because we over-index how good we are at what we're asking. We over-index our clarity or understanding what we want to actually ask it.

What kind of things are we not putting in silent space? Like, you know, it's a typing or ambient listening tool. It doesn't see facial expression. It doesn't know.

It hasn't probably guessed all your nuances of what you leave out, what you leave in the conversation. So you've kind of got to ask it and then realize you didn't ask it the right question or ask it more. The great thing is once you realize it's answering something you didn't, you don't feel bad about its feelings for making it do the work and then asking it to do it again. Because if someone else gives me the stuff, I'm like, I don't want to send them back to go do this again.

Oh my God, I sent them to do this and I gave them half-baked instructions and my instructions were bad. So the product's bad. Like that's, that makes me feel bad as a person. I don't feel bad about the AI tool doing all the same work over and over again to get it right because I'm giving it better instruction each time.

So I would say the bad is not clarifying everything that you want or adding to add more context to the discussion. Okay. And have you seen it just simply deliver, you know, factually incorrect information? I own a company in Texas and it gives me rates based on New York or.

I mean, I haven't seen it give a dollar amount. I've seen it cite the sources, and I've seen it give estimates in regions, and I've asked it to regionalize towards industry, region, competitor, like what's happening on those. And it gets pretty good in that. And it's pulling.

It generally tells you what sources it's pulling from. And I also ask it what sources it's pulling from to verify. And I have seen it give bad information, and it'll say like, oh, my bad. That was not right.

I gave you something wrong. I'm like, oh, well, thanks. That's not why I'm coming to you. Like, that's like, check this first, you know, and like double check your sources, commit to memory.

We don't have to have this conversation again. Those are things we have back and forth with AI these days. That's totally fine and doable. And I've, you know, in non, I mean, in non-insurance world, I now have two sockets in my garage.

I have a one and five sixteenths and a one and a half because Gemini told me the wrong size first. So I have the choice of whether I cherish this $17 socket or it becomes an Amazon return because Gemini was wrong on the size that I needed for my little Mini Cooper. I'm probably going to cherish the socket because I need it to complete my set. However, you know, it just gives wrong facts sometimes.

Okay, so what I'm hearing from you, though, is there's more good than bad? And I mean, accuracy. More often than not, you're getting accurate information. More often than not, I'm getting accurate information.

The clarifying questions have been a real game changer for me and my usage. Not feeling bad about asking for clarifying questions, having it do it three times, learning about myself. I mean, dear God, does my brain not work in a right to left up and down pattern? It does all sorts of 360 dive bomb circles as I do everything.

And so trying to get it to do what I want to do, you know, that is not its fault. It's not my fault either. Is it making you better at communicating with humans? To be determined.

Okay. Because you're definitely in the mad scientist category and like how your brain works. You skip steps two through seven. Yeah.

And if I'm silent, that probably means I'm choosing not to tell you all the different things I just thought about. So. Okay. We'll see a finite grummet.

I was going to say, what I'm now seeing with the extra steps are it being able to give strategy in a way that I really like. Now, it is giving some strategy that I might not want to go down with an employer because I already know other contextual clues about them. Though I'm having that moment of pausing going you know what let me not just discount it and take it off of one of these bullet points or tile boxes or whatever it's giving me to discuss with them after it takes my prompts and what I wanted to look at it's actually a good discussion to go through some and and Again, why should I, as the broker consultant, just take something off the table completely because I know it's not something the client wants?

I mean, there is a reason to take a finite amount of time and consolidate and focus on the things that we know we need to be focused on. But again, we can gloss over the things and still provide it because it's not taking any additional time to put that into there as long as it's not visual or junk clutter. But as long as it's relevant and we may just want to keep that in the vest this is possible we just know we don't want to do it like we can just leave it there and it lives in space and it didn't actually take the time to produce it because what we were looking for before is like what can we get done with the finite amount of time that a client's got to give to this area without bogging them down because i don't want to bog a client down i don't want to bog a prospect down like There needs to be an end result, and I want to get them there efficiently, fastest, graciously, and respectful of their situation that I can.

Now, does this give more context that I wouldn't spend an extra 15, 20, hour, hour and a half doing that they could if they wanted to take their own time, that I'm a bad judge of how much time they have? Absolutely. And so that's really kind of freaking cool. Yeah.

So how should employers and even their brokers really set that stage with AI being part of their benefit strategy, but I'll say specifically with their renewal? First and foremost, I think the broker should lead the charge. Okay. I think the broker should assume that that will be run through an AI tool.

They should probably run their own advice through the AI tool and see what the differences are and see their. Now, I also danger in this area because some brokers run really loose and fast and free. So, you know, broker warning, be careful what data you're giving it. Like, is this publicly available?

Like, is it a rate chart? If it's a renewal, like a dollar amount is not different. But sometimes we've got high claimant reports or incurred but not received billing reports and drug reports. Like, there are things we don't want out there.

You would repeat that to AI? There are ways to do it, ways not to do it. And so you want to be very compliant in how you do it. There's also ways to accidentally feed at things that you don't intend to.

A hidden column on an Excel sheet. I would so be guilty of that. It's possible, right? And so what kind of environment, how are you using tools, what protocols, how are you training your team becomes very important as we go down this rabbit hole.

Not a rabbit hole, a new pathway, a new journey. Yeah. And so, those areas. But I do think the broker should be leading the charge, one, in anticipating.

What the client's probably already doing but also using it as a tool to more robustly get that conversation going yeah now an employer should also be using and i think it should be an open conversation like it should be an expectation it doesn't need to be the i've got a little tool over here and it's going to help me negotiate with you or i've got a tool here because we're not very good at what we know or i want to try to steer you to this product and how do i get past these other things and nefariously try to just, you know, prove these other options, not nefarious, but someone wants is what they believe is the outcome.

But I think it needs to be an open, understood conversation that it's part of the conversation. We're using it. We're going to use it. We're going to get better at using it, both sides of the equation.

And I think that the more we can feed it and, you know, maybe even from what you and I are working on, on the podcast, the blog, the educational piece, all the super good stuff, the other research groups put out there, the more it learns from the source and it keeps learning from the source and we keep adding it to it, there's going to be a better, more robust conversation for everyone. And I'm seeing it come through, especially on the self-funded side with way more cohesive knowledge and ability to put thoughts together in strategic levers to pull on plans and how to do costs and how to negotiate and how to hold the vendors accountable.

Like it gives, it gives like check with your PBM on this, that, follow up with your TPA, think about this program, your population might need this going on is what we see. Like it's giving an action plan on a bunch of these things that I think empowers both the employer and the broker to start actually taking all the downline vendors that they don't really control, but they pay to hold their feet to the fire and get it done right. And so I do think this helps across the board and leverage it.

Talk about it. Talk about where we can use that, how we can have it keep on track when we're going to get sidetracked. Have it do check-in points. Is there a certain prompt or certain things that you would say if you were the employer to your favorite AI tool?

I mean, I would give it the here's who we are. Here's our experience. And I would go ahead and gather like your claims history experience, like at least like what is your loss ratios? What have your renewals been?

You know, are there high cost drug stuff you know about from those reports? You can feed it in a de-identified way that you can use it to say, what should we expect? What should we expect for an increase given our history? What are those strategies to use?

What are the conversations points we should do with our broker? What are the other opportunities? What other kind of plan design? Are available for our company makeup?

You know, give it stuff all the way around. Like, you know, are we thinking of having a merger, an acquisition in the next 12 to 24 months? Are we alone in our category? Are we a first, second generation company passing to another generation?

Like, how do these things in this financial structure play? And also, what are the pitfalls of these different insurance structures? You know, what does our plan look like if we stay on a name brand PPO versus going to some kind of, you know. Innovative or different structure like an open network or reference-based pricing structure?

What is the client experience? Is there a hospital pushback? How do we ask for disruption reports? How do we know these are going to work out?

How do we build good communication campaigns on the stuff we do choose to roll out? These are all very easy things to help us tackle. And I'm saying help us tackle. Because I think the other trap is we go ahead and believe that AI can do everything.

It's the instant gratification loop we just want to get it and be done with it we don't want to check it we don't want to refine it and we just want to get it and move on so we're trying to do even things even faster than we were already doing which does cause things that people call ai slop i mean ai does produce a lot of slop but if we're not watching it refining it teaching it teaching it our voice to get our company culture teaching it how we want to do you know branding for employee communications, all of those things, they take time, but they don't take an inordinate amount of time versus.

One, what we were doing before, or two, the dollars we were paying an outsourced vendor to do a bunch of the things we can now take on internally. And we've gotten back dollar resources for that time we're spending with people. Do you have a favorite tool that you think works better for these kind of questions? Or one that you would not recommend today, knowing things change quickly?

I mean, things change quickly. I was relatively disappointed in Gemini when I first started playing with it because we run on the Google Workspace platform and it was included with us. And my first thing was, well, what are we paying for? What's already in our environment?

What's not just out there in the ethosphere? Like, how do we keep things right here contained? what do we use, So I started working with it and didn't love it. I probably also wasn't as good at all of them at that point in time, so maybe it was a little bit of a bad assessment.

But I do like it. It's proven to me that it can do more and more. So that combined with Google's anti-gravity has been a fun tool. Google's Notebook LM product has been a really interesting one to use as like an educational learning format internally for our team.

I was never super into chat. I was a little bit kind of periphery on it. Started working with Claude a little bit as well and really found things that I was getting good traction on, really good traction on there. And I think that's probably going to change and flip-flop month by month, week by week.

And so, you know, with me, I'm thinking, you know, chat, Gemini, Claude, Copilot, you know, I haven't used Grok outside of the car experience, so I've never really tried it in that area. And each of them have their own deterrents or good things or bad things. And there's a ton of other tools out there and other tools that actually let you switch between the different language models, even inside of them. Like I mentioned, like Google's anti-gravity coding agentic tool has been really interesting.

And stuff that I don't do well, but it's also re-engaged my brain in that tech side that I haven't gotten to play with in a while. So for me, it's been a lot of learning by play. Oh, that's funny that it's re-engaged your brain. I actually have to take full-day breaks from it because I'm getting dumber the more I use it.

I used to code websites by HTML code and CSS code, and I haven't done that in 20 years. Okay. So seeing things that I haven't done kind of in that language has reengaged an area that my brain used to know a little bit about. Okay, fast forward two years.

What do you think the evolution of AI looks like in the benefits-buying discussion? Okay. Oof i am seeing i mean i have seen for the last couple years several software vendors pushing the ai piece there's one that we looked at that you know the pitch is we can replace your account managers with ai which yeah i mean so that that is the pitch like here or at least replace 90 of what your account manager does with the ai i've looked at it i think that lives in a very finite space.

I think that lives in the, the one to 21 to 50 employees, small group regulated ACA space that doesn't take into account underwriting. Like it, it can do things that are shelf rates. Like you live in Denton County, Texas. Those are set rates you can find by county by quarter.

Like those are just rates that are published and marry it with a census of ages for average age. Like That's just mathematical. So the AI can make the mathematical pieces in their fit. It can pull policies.

It can do the other things. It can do a bunch of those things. That's their sales pitch. It doesn't do everything else or have full scope.

But I mean, I do think that that's going to improve. So as software vendors... Get more money more investment more time in an industry that product will be there more and more so i do think that a lot of those functions that we would think of as an account manager function and let's be honest when you're after you're someone after a normal agency has sold your account your benefits and put a strategy in place the account manager does 95 to 100 of the work so if your account manager is now 90 in ai your experience is going to change very very much So I do think we will see a bunch of those functions taken over and you may not even know it as a consumer.

I do think that it will get into the self-funded space and the reports and the actions. And I think you'll have those tools handed to you as a consumer. So in two years, I think you'll be able to drive actionable strategic items from the employer HR admin end as efficiently as a broker can today. Wow.

So I do think this is a race to strategy and not so much of a race of function. Mm-hmm. Okay, in two years, I'm going to go back and listen to this episode and then either immediately delete it if you were wrong or be like, what is a - Hey, I have been so wrong before. In college, they wanted me to take computer science classes, and I might have even said out loud, the Internet is a fad.

I don't want to go down that road. I've been so drastically wrong and egotistical about what I know as little Brett in his space and time. I'm happy for it to be completely and abhorrently wrong in two years. You can leave it out there.

But if you're right, then I'm going to send it to everybody and be like, Brett called it. No, that's braggadocious and that's not me. I wish you had been right about the internet. I don't know.

We wouldn't have each other without it. We're in for an internet match. This is true. Why destroy a good thing?

Anything else you want to share? Anything we didn't touch on? I have been playing with the idea of sharing the prompts that I'm using, both publicly and with prospects and with clients. So if you're a client of mine, you may have seen, hey, here's this, but just plug this in for further context.

You might have seen if you're a prospect, I've done it a couple of times just on LinkedIn area. I'm toying with that to see what kind of feedback I'm getting. I'd love more feedback on it as people see it across the board. Okay, so people who hear this episode reach out to you at generousbenefits.

com. Yeah, well, I mean, if people hear this episode, always reach out to me at generousbenefits.com because, you know, I do run a company and I do need clients because clients make the world go round. Cool cool okay well thank you for all your insight um and for the delicious coffee absolutely if anyone really wants some delicious coffee it can be had if you asked yes brett bremmett expert in all things home roasted coffee and insurance ai and opal gts and maybe as we get to the future of the new project vintage mini coopers we'll see that was brett Brummett with Generous Benefits.

Brett's advice? Start asking questions, follow-up questions, and follow-ups to the follow-ups. AI is a tool to help further the conversation faster to get you to where you really want to go. Thank you for listening to the Generous Benefits podcast.