Anthony O'Donnell: I'm Anthony O'Donnell, and this is Life accelerated, a podcast for life insurers striving to achieve digital transformation. In this episode, we're discussing one of the most powerful tools for digital transformation today, artificial intelligence and its potential to significantly reshape how insurance companies operate. For this conversation, we welcome back Richard Wiedenbeck of Ameritas. Richard was actually the first ever guest on Life Accelerated for an episode published April 2022. At that time, he was the company's chief technology and transformation officer. Richard is an executive with tremendous experience in life insurance and other industries, and he was formerly CIO of Ameritas, and this year he became the company's chief AI officer. The existence of the title itself is an indication of the progress being made in the study and adoption of AI, both at emeritus and within the life insurance industry more broadly. Richard's new role provided a great opportunity to hear about how AI is enabling greater efficiency and smarter decision making at emeritus as the company navigates the challenges of cost and productivity in a rapidly changing world.
Anthony O'Donnell: In our conversation, we dive into the practical steps emeritus is taking to apply AI to fruitful use cases, to integrate it broadly into its operations, and to improve underwriting processes while enhancing productivity. Here's our conversation. Well, Richard, it's great to have you on Life Accelerated again.
Richard Wiedenbeck: Thanks, Anthony. Always a pleasure to spend time chatting with you.
Anthony O'Donnell: Yeah, absolutely. Back at you. So you and I recently had an exchange where you had mentioned that you'd been having some discussions about AI. This is obviously a hot topic, perhaps the hottest topic in many industries, and certainly life insurance. I wanted to ask, how is emeritus thinking about AI? What does the leadership of the company think about its potential and how important adoption might be?
Richard Wiedenbeck: That's a really great first question for us. We see the potential as huge. Part of asking me to step out of my CIO job and into a focus Daio job was part of us recognizing that potential. But our thought processes went around a couple of frames. One was, hey, we wanted to make sure that we didn't wake up three years from now and we didn't figure out how to take advantage of AI. So that was part of our thinking. We also didn't want to look at AI as just a tool. Right.
AI for AI, safe. We had a lot of conversations in the beginning about, we see this potential, how do we want to think about it and go after it, and then how do we want to structure and focus around it? And I think one of the things we looked at is, youve been following the life insurance industry for a long time. If you look at America specifically, weve done a pretty good job of top line growth. Right. We seem to do fairly well in continuing to grow ourselves. In the twelve years, 13 years I've been here, when I joined, we were a $1.4 billion company. We're over 3 billion in revenue. So we've clearly been able to grow, but we haven't really solved the cost curve challenge.
Right. Expenses continue to grow in step with revenues. And so we also took a step back and said, look, if productivity and efficiency is one of our main goals and objectives and focuses, and we don't have AI in our toolkit, we're missing a tool. So it wasn't a tool for tool, but it was like, hey, that's a pretty important tool to get in your tool bag. So that's probably a nice way to think about how we were thinking about it. We see that potential, but it's also, how do we want to take advantage of that potential in the way that helps us solve the challenges that we see on our horizons. Right?
Anthony O'Donnell: Yeah. So it seems like it's not so much this horribly mysterious technology that senior leadership knows nothing about. Rather, they're looking at it opportunistically as a tool that could be specifically useful. It sounds like theres an idea that it could be an aid to automate processes that are now costly and that you havent been able to reduce that cost. But let me ask specifically what measures emeritus has taken to study AI and investigate its potential. It seems like youre maybe not in, as a rudimentary state as some others might be, but what have you done to better understand AI and its potential for reducing costs?
Richard Wiedenbeck: Yeah. And I look, I think like all of us were, we want to make sure we, we become smarter, right? That we don't just assume we live in, you know, it's. Very few people can claim high ground in this space. Right. Stay on top of the research, read what others are doing. We formed a book club as an executive team because a lot of this is a mental model shift for a lot of executives. I look at AI a lot like, and I know for us that have been around the technology front for a while, it's got a little bit of a feel like the Internet. Remember when the Internet was first coming out, you know?
Anthony O'Donnell: Oh, absolutely, yes.
Richard Wiedenbeck: What we thought about the Internet on this side of the Internet versus what we realized once we were in it. So part of what we said to get smarter and study AI, investigate it, is we've got to get in it and learn while we're in it, because we'll make smarter decisions while we're in it. Right. At the same time, life insurance is a heavily regulated industry. And even if it wasn't, we want to make sure we're thinking about what is our responsible framework for governing where do we want to use it? Where do we not? So I would say the measures we're taken to study it and investigate its potential are both on the risk side and the value side, and then building our frame of where do we want those diapos to be? Right. Which kind of ties back to my first point. It is then easier to start on the operational efficiency of productivity area because you actually have less to navigate in terms of risk to your customers, risk to your products, risk to regulators. They're way more focused on, are you using it to make underwriting decisions? Not that we haven't been using selection and criteria to make underwriting decisions, but throwing AI at it starts to get everybody's hairs up.
So it's like, all right, no, I have to go there first. Let's go learn over here and then come over there. So I think it's that journey of learning, experiential learning. In this case, I don't think you can get there by just reading something. I think you've got to get your hands dirty, your feet wet, but do it in some zones where maybe the risk of the firm and the risk to your ability to stay true to your standards is not as high.
Anthony O'Donnell: Yeah, I was thinking as you compared it to the Internet, that obviously there were some very wild speculations about what it was going to mean, and it was supposed to disintermediate agents and direct selling was all the rage. That was the way the Internet seemed to be likely to be used, which turned out to be only partially true, turned out to be a tremendously valuable tool for carriers to interact with agents. I thought maybe it's also like the Internet in that it's likely to become a really pervasive technology in that regard. Maybe more like analytics, and maybe it's like, you know, what we used to call about, what was it? Olap, online analytical processing, and likely to be so embedded in systems that it becomes pervasive. Is that the way your leadership is looking at it?
Richard Wiedenbeck: I think, and this is probably one of those where it's like, look, I have my version of the future, I think that will occur. And I'm kinda pretty sure I'm probably half right, half wrong. And so I think it's also, how do you navigate that? I think when we see the potential, we see a full range of potentials. Our desire to go after all of those potentials at once is tempered. And we've said, look, I want to go after the operational and service deficiencies because we believe that use the tech to solve a challenge and learn from that, and then turn around and then start walking into these other spaces. Right. So thinking about can it help us understand mortality risk better, maybe we're not going to start there. Can we use it to understand where our experience is moving off of our pricing assumptions? Absolutely right.
Help our product teams be smarter about the decisions we're making. Help our understanding of our mortality risk relative to our experience be faster and smarter. Yes. Turning that into how we go design a product differently. Maybe not so fast. I think the potential is there. But the other thing I would say is AI is not new to us in a very targeted way. We've been using AI to read dental x rays and help our dental claims process for five, six years.
I mean, we've been doing that for a while. We've been using it to help us have machine learning models around our experience, studies where we've been using AI in very pointed ways. I think the difference is, Jeremy, AI is like this uber moment where your ability to bring it all together, and then to me, the third piece of that, which we are just now trying to dab our finger to, is what I call agentic AI. And that is having AI agent work like a human. I still got some old systems because there, how many policies do I have that are 80, 90 years old? And it's hard to train people on those systems. It's doing, I can teach an AI agent to do that work. Right. Those systems are closed blocks.
Those screens don't change a lot. That tech isn't updated a lot. It's hard to train people on it or keep them excited about going in and getting f four and f six and typing it in. I can get AI agents that they don't care. An AI agent doesn't care what green screen or pretty screen or beautiful experience. It's just going to do it. And so there's some neat places like that where we can just turn it loose and let it go do work. But I think there's where the big potential is, can I take an apartment that has 20 people and those 20 people can now work like 60 people? Boy, that's going to help me not have to add cost to the structure as I grow.
Right? If that makes sense. And I mean, if you want to go deeper, just let me know. There's probably other places you want to go there.
Anthony O'Donnell: Well, what I wanted to do was kind of move from the abstract to the concrete. You're talking about these potential uses. You mentioned earlier, you talked about getting your hands dirty and your feet wet will give you ten demerits for mixing your metaphors. But are there examples then of how Ameritas is actually moving from the abstract to the concrete? From speculating about what might be useful to starting to experiment?
Richard Wiedenbeck: Yeah, and I think you hit it right. We see this journey kind of in some stages that may not be as time based as they are event based, but so stage one is experimental learning. And they're simple use cases because that helps us learn. So our investment area. Right. The number of ten ks they have to read. AI is great for jumping in, running through those ten ks, reading multiple pages, summarizing it, sending it back, giving them the five things they're looking for, and let them move on. So there's a general productivity that we can get.
Our underwriters, they have to look at attending position statements that are handwritten, some documented, some scribbled, some whatever, and the AI can weed through pages and pages of that. Here's the five things I'm looking for. Here's the questions I have. Not only give me the answer, tell me where you found it, on what page, on what docs, so I can put that in my underwriting decision. Those are real cases that we're working on now. A lot of them are summarizing, searching through large amounts of documents. Those are kind of the early, easy use cases you can go after. I'm getting multiple rfps in multiple formats.
Can you read through that RFP, find the ten salient points, help you build a quick response where it would take us 20 days to reply to an RFP, now takes us three. But they're simpler use cases. Those are the kinds of things we're trying to learn by, because they're time consuming tasks. They're tasks that just require somebody to weed through lots of documents. And the AI is really good at that. You could handle all of that and then walk through it really quickly. And we see that as stage one of our experiential learning we're dabbling in. Can we now go do more complex use cases? So we want to move from stage one to stage two.
Pretty good, which is, let's get a case that might have some of that work, but I want to string a little of that work together and add a little workflow around it. So now, once that's done, can you hand it over here and go do that next piece of work? Because now you got a little bit better productivity and efficiency gain off of that. But we're very early stage there. I would say those experiential ones I gave you are good examples. And we're seeing other firms. Same thing. Large volumes of data, easy to summarize. We are looking at hand at these 29 spreadsheets and have it summarized.
The ten things that it's looking for, the trend that it's seeing, or the data. We're still in early stages. Some of those we're seeing work well, some of those who are not seeing work well. So we're trying to understand why. Why did you not get it right? Was it just, it was hard to navigate the data? Here's a really good example of getting the information that you want your AI to use right. So we had this really interesting conversation on how do we help our call center agents answer a question? Everybody likes that one. And the first thought was, well, let's go get all the call logs from the last five years, and let's ingest those, and then it will know. We answered the question the last five times like this.
What we uncovered was five times out of ten, you answered the question right. The other five times, you answered the question wrong. And you're asking the AI to discern between which time you answered it right, which time you answered it wrong. And that was because the information set you were given, the AI had both right and wrong answers in it. And what you're trying to do is help a CSR answer the question in the right way. And what we found was call logs were the wrong source. The policy docs and the admin memos were the right source. So I think there's a little bit of learning on what's the right source to give your AI, so that your AI is doing what you actually want it to do.
And you're not asking your AI to traverse the because you're asking it to discern which answer is the right answer. And because I said it seven times, does it make up the right answer you could have? Right. You know, I got a poorly trained CSR, they answered it wrong, seven and right three. And the AI says, well, the last seven times you said this. So let me just keep propagating the wrong answer. But those are really good examples of putting it to use, getting in, digging into the practicality of it, and then learning that learning was healthy. We learned, hey, that's probably the wrong source. Let's go to this source.
So you've got to constantly be paying attention to what you're getting in those early stages to make sure that you're getting the result you're after or the effects you're after. And that wasn't really biased. People talk about bias in the models. It was like the bias was in your information set. You gave up right and wrong answers. So those are some examples, real examples of what's happening. I think. And I hear other life insurance companies probably doing similar things.
Anthony O'Donnell: Yeah, yeah. And it's interesting too, because I think in the examples you've given us so far, we've seen some focus on efficiency, but also creativity in approaching certain problems. And it's very interesting to see that a major challenge of AI is just the training. You said the source of information, and I think thats part of that. How do you train your AI? What do you give it to? Process. And that must be part of it. I wonder whether youre doing any thinking about how it can be used on the sales side and illustrations or whatever the old saying, life insurance is sold, not bought. Might there be creative ways to boost growth through better approaches to sales?
Richard Wiedenbeck: Yeah, and I think this was where we may be unique because I think other life insurance carriers are clearly focused on going after some of that. We're a little more hesitant to jump there to start. I think we want to be a little more insight focused to where we're smarter. We understand how to use the tech, we understand where it can be right and where it can be wrong. I would say we're trying to go up. It's a similar use case, but can we help ourselves? Support areas, find the right documentation, find the thing quicker to give it to an agent, but we're really, at this point, other than that RFP scenario in our dental division, we're really not right now going after that kind of in the sales revenue. 85% of our use cases are going to be efficiency. I don't want to ignore a good revenue generating opportunity.
Right. So we say 15 there and then five and just make life better because we know that's always the greater good and we're just being a little more methodical and hesitant about jumping into the revenue side.
Anthony O'Donnell: Yeah, well, and also it's not as much of a problem for you as it would be for other carriers, given the figures you discussed earlier.
Richard Wiedenbeck: Yeah, I mean, I think it's, it doesn't mean there's enough places we can't go help there. But I think, yeah, you're right. It's like, look, there's a bigger challenge over here. Let's go solve the challenge that's on our pocket when I think those will reveal themselves. Right. I think those opportunities will. And back to my Internet comment. As our people are smarter, I think sales side of our company, the distribution side of our company, if we get them smart about the efficiency of productivity and use of AI, they'll start bringing forward use cases to help with revenue.
Right. That's what they get up and do every day. And I think they'll start saying, so instead of me trying to push it on them. I think we'll create a pull effect there, hopefully. If not, okay, we can go focus on that later. But it's just not been the big push for us. We've been more on the other side.
Anthony O'Donnell: And in the meantime, you're learning a lot about what you can do with.
Richard Wiedenbeck: This technology and that is, I think, a really key part for us right now. Yeah.
Anthony O'Donnell: So I wonder, Richard, where vendors might come in on this subject. Like you're talking a lot about internal use. Are vendors already in that mix or are you vetting vendors for certain purposes with a look to their ability to use?
Richard Wiedenbeck: Yeah, vendors. Can I sigh when I say the word vendors?
Anthony O'Donnell: Vendors are our friends.
Richard Wiedenbeck: Yes, they are. And they're well intended. Right now, there's a little bit of a version of crazy out there with vendors. It's like every vendor is walking in the door slapping an AI logo on their product. And in all honesty, if it was a feature function improvement set, and it was, you have my product and now it's AI enabled here, that would be great. But all of them are using it as additional cost mechanism and additional. And that potentially is a good thing for us because we do have to go figure out where should I use AI embedded in a product and where should I use my gen AI tool and where should I use agentic agents and where, quite frankly, should I use robots, good old fashioned automation. At the end of the day, if we're going again after process efficiency and productivity, there are other ways to solve that problem than just AI and we don't want to lose sight of that.
I think we're also early stages on Microsoft 365 as a copilot. And when do you use the copilot inside and when do you go use our Gen AI tool and we're trying to figure out there's a big user challenge there. How do I make sure you're using the right tool for the right job in the right environment?
Anthony O'Donnell: Yeah. And a governance challenge.
Richard Wiedenbeck: Yeah. And everybody is selling us their thing, right. Everybody's also selling us gen AI, baked into it and you're like, hey, Salesforce, I love you, you've got a platform, but I'm not sure you're OpenAI or you're Amazon and are you gonna step into that space as well? But we are seeing a lot of players, I would say enter into spaces. I think it's going to take a while. It's going to shuffle itself down. It's going to get there over time. So I don't know if I have a magical answer. It is one we are just continually navigating and we're starting simple like I am, and we have strong AI governance.
You cannot turn AI on in a system without that going through good old fashioned technology, due diligence with our risk folks involved or security boats involved, our legal boats involved. We have some AI overlays, both on what you call the traditional side of security and risk and on the what do we want it to do? So we actually have not just what it can do, what do we want it to do? Overlay. And if we're like, we're not interested in using it for that right now, then we're not going to let that tool come in or we're going to do our best to not have that tool go that way. But it's not easy. I love the question. It is probably the most challenging area of AiH because some of those things, vendors are just, some of them are just turning it on and now your users are using it and we haven't had a chance to bet. Hey, your AI is using an open ended LLM in my call centers knowledge management system where people are used to docs that they can put sensitive information into where's that going? And we don't want that going somewhere. And that vendor didn't tell us.
I don't have that control on it. So we're really trying to make sure we, as much as I hate to say it, tighten those hatches down a little tighter right now so we can get smarter, because I think they're bringing it in. But you got a bet. What did you do to me when you brought that in? And that's creating a very interesting dynamic there. I don't know if I have the perfect answer other than we've got some very strong governance and tech due diligence around it. And it may feel a little constraining to folks right now, but I think it's smarter to turn the screws in the beginning and then open them up as we get smarter than open them up. And the law of unintended consequences can move very fast in this space. So I think we, again, we're heavily regulated industry and a risk of purchase industry.
So those two behaviors match us really well, but we like those two behaviors anyway. That's how we're going after that.
Anthony O'Donnell: Yeah. Well, I'm glad you liked the question, and I thought the answer was rich with content, as they say these days. One of the things I liked about the answer was the implicit good advice for vendors. And I wonder whether we might conclude with what advice you might have for other senior executives at carriers when it comes to the potential of the technology for the insurance industry.
Richard Wiedenbeck: Yeah. And, you know, I'll go back to a comment I made earlier. Right. I don't think I can't claim high ground on this. Neither can anybody else. Right. There's anybody who stands up and says I'm a 30 year AI expert is lying to you. We're all kind of learning.
But what I would say is what a great opportunity to have a meaningful impact on your company. We've chosen where we're pointing it. But again, this is a technology with huge opportunity. So I would say what a great opportunity. Don't be afraid to lean in and go try to figure that opportunity out of. I think you have to. No matter how you go after the opportunity, I think you have to say, how do I make sure I have a responsible governance frame that decides what I do and don't want to do? I think on top of that, you better have a good education, training and upskilling program because the way you're going to do work is going to change. And how do you navigate your workforce so that they can now do work in a new way, they can interact with this in a new way.
So to me, that's got to go along with it. But I think at the end, it's, don't run from the opportunity, run to the opportunity, because I think it's a run. Run it in a thoughtful and purposeful way, but run to it. And if somebody's asking you to take this on, I'd say, God, jump on it. How many times in any of our careers do we get an exciting opportunity to work on what I will argue truly is a game changing technology? People have put this in the fourth industrial revolution. It is. I think. Yeah, I think it's going to change a lot of industries, not just one.
Anthony O'Donnell: Yeah, you focused on opportunity here, but if it is a game changing technology, then there's a threat as well.
Richard Wiedenbeck: Oh, absolutely.
Anthony O'Donnell: That if you ignore this technology, you're going to get left behind.
Richard Wiedenbeck: Yeah, I think both sides of that coin are true. Right. And how fast and how quick and human adoption and regulation and all of those are going to play. But to your point, sitting on the sidelines also, it's like, mike, comment on not having it in your tool bag feels like a bad answer. Sitting on the sidelines feels like a bad answer, too. Yeah, you gotta get in the game. And then, like we said on the Internet, on this side, you didn't know what you're get in and start to make your decisions from a position of being well informed and understanding and the real experience, and then navigate that cost, risk, value curve the way you wanna navigate it. Right.
And I think that seems pie in the sky kind of stuff, but I think it's real here and it's exciting, you know, I mean, it has re energized me, and I'm in the final stages of my career in emeritus. But they didn't look at me and say, hey, we're gonna put you out the pasture. They came over and said, hey, we want you to kind of make sure we're in a great spot here and your last two to three years at this firm. And I would just tell people, again, run to it, don't run from it.
Anthony O'Donnell: Sounds like good advice to me. Richard, thank you so much for sharing your insights here on Life Accelerated.
Richard Wiedenbeck: Yeah, no, always a pleasure, Anthony, to spend time chatting with you. You accommodate these subjects really well and early and often. And I'm glad we had a chance to reconnect, you know, glad the opportunity for me to get into AI aligned up with kind of the stuff you're focused on. So again, always a pleasure and thanks for having me on.
Anthony O'Donnell: Thank you, Richard. I think listeners would agree that hearing Richard Wiedenbeck talk about the potential of AI at emeritus is not likely to be comforting to executives wedded to a cautious look and see approach. But then, this is a man who during the last six years has run 84 half marathons in all 50 states in the District of Columbia and is scheduled to run the full New York marathon in November 1. Of the things that the conversation clarified is that AI is playing a transformative role at emeritus. The technology isn't just about automation, it's about smarter decision making and creating efficiencies that can support growth without escalating costs. While emeritus is focusing more on the back office first, the company is quickly gathering vital learning about AI to apply across the life insurance value chain. To find out more about advancing your own digital transformation, check out equisoft.com. Life Accelerated.
