AI Doesn't Know How to Do Your Job. You Do.

Paul Tyler: Hi, this is Paul Tyler, and welcome to another episode of the L&A Hub Podcast. Today I have a really interesting guest — someone who started in the insurance industry and is now working at an AI company. Jane, welcome to the show. I want to introduce Jane Vevea. Jane, can you tell people who you are and what you do?

Jane Vevea: Twenty-year life insurance lady. That's what I always tell people. I started in this business at Prudential in 2006 in retention, training, and wholesaling. Then I moved into the bank channel at PNC, and then got laid off. That was a very interesting time in my career. I decided to buck the system and go into insurance tech. I moved to White Swan, where we were sort of on the precipice of AI, but not quite there yet. Then I moved to Atidot — which is Hebrew for "fortune teller" — which truly was AI. We were predicting lapses, surrenders, upsells, cross-sells — anything for your inforce block. And now I'm at xAI, working on the large language model as a finance expert.

Paul Tyler: You broke the mold. It's interesting thinking about your career arc — honestly, every time you're okay at a place, that's not necessarily a good thing. You sit and do the same thing over and over and don't get out. Sometimes a layoff, a change, something abrupt actually opens crazy interesting doors. How did you feel? You'd done call centers, sold insurance, wholesaled — and then this sharp left turn. A lot of fear, or more like "I've found my calling"?

Jane Vevea: Really glad you asked, because I've never gotten to elaborate on this. I was a free lunch, welfare kid. I grew up really poor, so my entry into life insurance wasn't driven by passion or calling — it was necessity. My sister said, "Hey, there's this call center job at Prudential paying exactly what you're making right now, working ninety hours a week running a Caribou Coffee shop." Thirty-two and a half hours for the same pay — let's go. That's how I got into life insurance. I didn't even take math in college. The fact that I got my CLU is already left field for me — the math made me cry. I stopped at Algebra One in high school. I probably shouldn't be announcing that at the top of my lungs, but now we have AI, so the math part is solid.

Paul Tyler: Ha — that's crazy. I think there's a lot of fear out there. I've been at three or four different events with people at various levels of engagement with AI, and people are confused. But I think, like you're saying, it's leveling the field in a lot of ways, if people take the time to learn these tools.

Jane Vevea: It really is. And it's actually exposing those of us who were sort of quietly — I wouldn't say ignored, because I'm too loud and too tall to ever be ignored — but those of us whose soft skills were undervalued. My creativity, my ability to build a PowerPoint was never prized as much as the ability to crank out emails or get forms done. Those were the "harder" skills. But those harder skills are exactly what AI can now do. The things AI can't do are the things I went to college for: creativity, ideation, the human side of life. This is a really interesting time for me to be working.

Paul Tyler: It really is. I knew a crowd at Atidot — Dror and the group — and I thought what they were doing was really good. Retention is a fascinating problem for insurance companies because you put this policy on the books, but the big assumption is how long it stays. If you're off, it's costly to long-term value. I want to ask you — what was the easiest part of getting companies to look differently at their inforce block, and what was surprisingly hard?

Jane Vevea: What was eye-opening was that when I was going to carriers in 2023, they still hadn't even tackled retention. They'd been writing off lapses as a marketing expense or saying it was already built into the actuarial calculation. It was just never something they cared about. I worked at Target when they got rid of plastic bags, and I thought it was a racket — they were charging for paper bags while the plastic bags you weren't getting were already priced into every item. When I went to work for Atidot, I thought about it the same way. Why wouldn't you be double-dipping? That opened up all the legal discussions and the ethics discussions nobody talks about: when you know someone is going to lapse, what obligation do you have to act on that? Those are discussions that haven't been deeply explored because we've never had AI. I never thought about how you'd pay attention to one group and ignore another. I thought: if the machine tells you these people need attention, just give them attention. Isn't that what everyone wants? Like Amazon — I was talking to my husband about camping lamps, my phone overheard it, and suddenly I'm being shown the best camping lamps. That removes all the inertia that prevents you from buying. It's just easy.

Paul Tyler: So who in an organization wanted to capture that value — make more money, keep the business — and who said "don't bother"?

Jane Vevea: The easiest door openers were always marketing and sales. They always want to make more money, and they want different ways to make it. In sales it resonates really well to talk about retention because you have this untapped resource, and you have a dying agency force you need to attract agents to. You can't just say "go talk to two hundred friends and family" — we know that doesn't work. So giving them pre-made conversations that are accurate to the minute — about why and when and everything we know about these consumers — makes that conversation more human. Like: "Hey Paul, I'm giving you a call. You've been a twenty-five-year customer with XYZ life insurance company. Our AI system says today might be a hard day for you, and I just want to check — is that what's going on? I can't affect your past, but today is a day I can affect your future, and I'm here to help if you need it." Then you have that human discussion.

Paul Tyler: I found that operations loved conservation work, marketing and sales loved it. The actuaries were always like, "I've assumed the following." But life insurance is a complicated instrument, and it's interesting how it got sold. We finally got the analysis done at my last company, and the biggest predictor of lapse — you can't make this one up — was whether the person called back and wanted to talk to us within a month. It was sold to some people who'd never had insurance before, and as it turned out the biggest problem on those calls was whether the premium deduction lined up with payday, because they were living that close to their paycheck. Who would ever think of that until you dive into the details?

Jane Vevea: Exactly.

Paul Tyler: And there's a flip side — cases where we actually wanted certain people to lapse. That trips right over into the ethics.

Jane Vevea: Yes — and those are conversations you can't have on a recorded line. Like: "Hey, we got this policy from the other company." All that consolidation from the eighties and nineties, all those different flavors of policy they'd love to get rid of — that's all sitting there.

Paul Tyler: I remember Drew showing me an early interface — this was radical in 2016 or 2017 — a dashboard where you could load in data and actually pick the model to run it against. That was radical. But I didn't have anyone at the carrier who knew how to use it.

Jane Vevea: By the time I finished there, we had the whole thing packaged end to end. You give us your approved marketing materials and we automatically send them out. You don't have to do anything but sit back and either get a hit on your website to a chatbot, route it to a person, however you want — it's a complete circle. I was amazed every day. But it's life insurance — we couldn't even look at Facebook until it had been out fifteen years, let alone market or sell on it. We're slow to change. But where we are seeing it used — inside underwriting, inside these operational workflows, taking the application cycle from sixty or ninety days down to a minute or two — I'd take four days at this point.

Paul Tyler: From my vantage point, I still think we need more people inside these companies who know how to use this technology in order to absorb it. Do you think that's changed? Tell people what you were doing in the early days of ChatGPT.

Jane Vevea: In the early days it was writing emails, helping with resumes — that's where I was too. But I think you're right, and I think the worst thing a company can do is take a top-down approach with AI. Somebody else said this — I won't take credit — but AI doesn't know how to do your job. You know how to do your job. And you know which parts of your job are terrible and annoying and you want to automate. It's so simple now to just talk in plain language and say: this is my problem. I'm tired of entering the client's name in the same form five times. I've had to run twenty illustrations and it's always these five — can you create an automation that runs them automatically when I give you the client name and policy number? So I don't have to keep doing these same boring jobs no one wants to do. And then I can actually pick up the phone and talk to the client about how their family is doing. I can be more human, and I can have the conversations that make life insurance the tool it's supposed to be, instead of the one that sits in the back drawer that no one thinks about.

Paul Tyler: I couldn't agree more. But it's rare, Jane. What you just described — here's the specific problem inside an insurance company, here's how AI solves it — is still a rare conversation. We still have people on one end who know the problems but not the tech, and people on the other end who know the tech but not the problems. If you were advising a carrier, running their AI strategy — what would you do?

Jane Vevea: First thing: a closed internal AI. A large language model the company uses only internally so nothing leaks out. You wall it off — and inside, it's like your best tenured employee. You give it all the knowledge, all the procedures, all the institutional memory. Like that one person who'd been at the company forty years and you definitely did not want to retire, because she knew the most obscure stuff. That's who you want it to be. Then you want champions in each area — somebody who's like me, genuinely interested in getting more efficient, annoyed by the monotony. You sit them with someone who is tech-native, because once they jump in the pool, they're going to figure out how to swim. It's just that first step. And once those champions are comfortable, they disseminate it down to the rest of the team. They say: "Here's what I did. The boss saw it, approved it, it's okay — and it takes five minutes." All of a sudden you have productivity in one day. There will be setbacks — you gain here, you break something there — but that's how every improvement in life insurance has worked, from the Xerox machine to PDFs. Everyone hated scanning documents at first.

Paul Tyler: The pool analogy — it keeps getting deeper. I struggle to keep up. Watching YouTube videos, back on X, going to Reddit — it's a challenge. Agentic AI has been talked about for two years, but now it feels real. How do you describe agentic AI versus what we think of as ChatGPT or Claude?

Jane Vevea: Okay, this is a good one. Large language models are just words. ChatGPT, Claude, Grok — they're a collection of words scraped from the internet and books. That's it. I say this as someone who has been in AI for almost five years and still feels like a total imposter. Agentic AI, to me, is more like automating those annoying jobs. What's the most annoying part of being a life insurance salesperson? The illustration tool, figuring out which policy to pick, filling out the application, following up, placing the policy. Agentic AI does all of that. All I have to do is find the client I want to talk to and say: I'm a life insurance person, I know I've got the right product. Do you like me? Do you trust me? Should we do business together to protect your family? Great. And then everything else runs in the background. I make money, my client is happy, and all I have to be is a human being who cares about protecting families.

Paul Tyler: I've heard it said that by 2027, everyone will have their own personal agent. Should every insurance agent have their own personal agent by 2027?

Jane Vevea: I think they should have their own agent that they interface with, but that does not interface with the public. That's a necessity for anyone in any job. Gone are the days of the standard operating procedure manual. In its place: an agent AI that holds all the knowledge, all the procedures, all the taxonomy you need to do your job — and as it improves, it starts taking over more and more of those tedious tasks. I don't think we're going to get to full agentic AI in life insurance anytime soon, and frankly we don't want to. We don't want it making final decisions. A lot of what's coming out of regulators is exactly that: use it to inform decisions, aid decisions, but keep a human as the final check. I'm on board with that. Let's use it to read a lot, but double-check its work. I have not found it to be consistently accurate, and I think we need to own that.

Paul Tyler: I've been playing around with voice agents. I won't mention company names, but you know who they are. I've been trying to do what you described on a sales desk — take an inbound illustration request, have the agent gather all the client info, figure out what they really want, suggest alternatives. Super time consuming today. Could an agent run that workflow and come back to me with: Paul, here's the plan? That would make me more efficient. Do you think tools like this will reduce staffing on a wholesaling desk, or does it mean the organization can do a lot more for agents than they ever could before?

Jane Vevea: From a wholesaler perspective, what I'd really love is for a tool that surfaces the right product without me having to brag on it or buy anyone dinner — because the product stands on its own merits, based on what's actually in the client's best interest. I believe deeply in a fiduciary responsibility. I'd love to push the entire industry to fiduciary, especially in the age of AI. If we can have these tools perform that suitability or fiduciary compliance layer — ensuring people aren't selling something because they get paid more for it, or because the wholesaler on the desk only knows one product — that's where this gets powerful. When there's a table shave at a carrier, you had to know the guy who knew the guy. It was never like: "Prudential has a table shave today, here's the notification." That insider information advantage is what these tools can democratize. And the products are changing as fast as AI is. How does an advisor stay current on life insurance plus annuities plus investments? That's too much. These tools can help advisors give better answers with more confidence — and open markets for insurance in places like the RIA space where people know they need protection but don't feel confident enough to act.

Paul Tyler: I think done right, these tools should make advisors provide better answers and give them more confidence. But let me shift to something else — are we sometimes over-trusting? We own PolicyGenius, and leads coming through ChatGPT convert at an incredibly high rate. The bot told them to do something, so they do it. When is that a good thing, and when is that a problem?

Jane Vevea: That's a really interesting question I hadn't heard before. I find myself doing that too — I just bought a pop-up camper and I was doing that exact thing with campsite searches. I think it comes down to how the recommendation is served. If the client is genuinely seeking information and the tool is surfacing what's actually in their interest, that's great. Where it gets dirty is in the context. The clearest bad example: if someone is in a mental health crisis and you're serving them life insurance ads because your algorithm flagged a conversation about suicide — that's obviously wrong. Black and white. But there are a lot of gray versions of that. If it's dropping in prices after an unrelated conversation, that's creepy. If it's entering a conversation the person initiated and genuinely wants help with, that's different.

Paul Tyler: Marketers will overuse it, it'll get cluttered, and then it won't work as well — that's predictable. But there's something psychological about trusting the machine. I get in my car, I don't check the battery every day. I just press the button. If it worked three times, I trust it the fourth time without looking.

Jane Vevea: I find myself doing this too. But I'm an unusual case — I was always a foreigner in life insurance, and being a woman in this industry means I always had to cite my sources. So when I use AI, I tell it to cite its sources and I actually go look at what it's citing. Most people I know are trusting whatever the algorithm serves them on social media, trusting whatever the large language model says. That's a real problem. And at work, you get people who become so over-reliant they're not doing their job anymore — "the machine did it." We're paying you to review what the machine did. So do your job, or find a different job. Accountability has to be built into how we use these tools. And honestly, I'm not optimistic that people will hold themselves to that standard unprompted. My dad sends me AI slop daily — my sister thought I was being mean when I told him it wasn't real. I don't think he gets it.

Paul Tyler: I've been getting people responding to my messages with a link to a ChatGPT conversation. Like — here, just read this.

Jane Vevea: That's wild. I could see it for a specific work purpose — showing where the tool made a mistake, asking where the logic broke down — but not just "here, read this" with no context.

Paul Tyler: Yeah, "give me feedback on this." I'm not doing that. Crystal ball: what do you feel good about, and what concerns you? Try to keep it focused on insurance.

Jane Vevea: What's great: the agentic agent for the advisor. What you're working on, Paul — I know a lot of people are working on this — the idea that we can take those tedious tasks off the advisor's plate entirely. And the pace of change is exciting too. Even if what the tool does for you today is just fill in names on a form, next week it might complete the whole form. Once you're in the water, the improvement compounds. If you're someone who finds that efficiency and that additional knowledge helpful, you're going to do really well. If you're resistant — you won't be in a good place.

What scares me: the inherent bias. We're already seeing research showing that when AI writes a resume with a male name, hiring managers see it as innovative. When the same resume has a female name, they call it cheating and don't even read it. As a woman who loves AI and is now being viewed through that lens — again — I hope we can learn from past mistakes. But my real fear is that the deeply embedded biases in our industry just get amplified by these tools. When we look at who runs this industry — it is still a very male-dominated executive suite, one of the last holdouts — if we use AI to look to the future by pointing it at the past, we get garbage in, garbage out. We have to give it our aspirations, not our history. Only then will we get out of it what we hope for.

Paul Tyler: Jane, this was great — I could talk a lot longer. For people who want to learn more about what you're doing, what's the best way to reach you?

Jane Vevea: Hit me on LinkedIn, or email me directly at janev at gmail. LinkedIn also has my phone number — you can text me. I've gone from old school to new school: I'd rather have a text than a phone call at this point.

Paul Tyler: All right. Thanks for your time, Jane. And thanks to all our listeners — join us next time for another great episode of the L&A Hub.