Insider Insights Podcast — AI in Insurance: A Level Set
Kartik Sakthivel, chief information officer, LIMRA and LOMA and Paul Tyler, chief marketing officer at Nassau Financial Group, level set the state of AI in the insurance industry. This podcast was recorded while Paul Tyler was CMO at Nassau Financial Group, he has since transitioned to a new role at Zinnia. Read the Transcript.
Transcript
(Auto Generated)
Welcome to Insider Insights, where we dive into hot topics facing the financial services industry. Today, Kartik Sakthivel, LIMRA and LOMA’s Chief Information Officer, and Paul Tyler, Chief Marketing Officer at Nassau Financial Group, level set the state of AI in the insurance industry.
I am super excited to be here with my friend, Paul Tyler. Paul, I'm super excited because, like I said to you, you're probably one of the smartest, most pragmatic CMOs that I know. I am really excited to talk to you about AI, specifically AI in insurance and where we are, and where we've been. How are you doing today?
I'm great, Kartik. And, you know, to invite me onto a podcast to talk about AI, oh, we could a three hour show on this. Maybe give, one of these other podcasters a run for the money, but I won't do that.
You know what, Paul? If you and I start talking AI, we could probably continue for many, many weeks, and it's not just three hours. So, I know, just curious from your perspective — you've been in the industry for a while. AI in our industry is not new. Right, Paul? We've been doing AI across the value chain for a good long while. Why is it that AI has occupied, our hearts and minds as much as it has over the past twelve to fifteen months?
Yeah. It's interesting. Yeah. Twelve to fifteen months. I think ChatGPT launched the end of November 2022.
You think about that. That's I don't know — depending on when the show comes up — maybe five hundred and eighty days. And look, it fills my inbox with emails, and on television.
Yeah. So I think this technology is one of these waves of technology where, if you get into it and understand it, you can see how many parts of our business can be dramatically impacted for the better. And I kind of liken it to a key you put in a door to open up some technology that has been sitting here forever. As you mentioned, AI has been in our insurance space for a number of years, you know, with predictive underwriting, a lot of analytics behind it. But it was it was all kind of you know, you had to be a data scientist. You had to be a PhD. You know, you had to have been a rock star programmer to actually put any of this stuff into production. Now suddenly, you know, carriers can do this — it's something we use.
Yeah. I think the cost of entry for generative AI has made it so easy. Right? It's effectively democratized artificial intelligence to the masses. Right?
Yes.
The other thing that, you know, you've heard me say this before. I do fundamentally believe that any technological advancement that actually takes root in our personal lives and then goes into our professional lives has a lot more staying power. Right? Which is why, hopefully, no one out there is using a Blackberry. I mean, we all loved our Blackberrys, but let's face it. Right? It was, it was completely disrupted. I think that's number one.
Number two, generative AI. I mean, you mentioned ChatGPT, Paul. I remember you and I having a conversation about ChatGPT sometime in 2023. What did it take five days to reach a million users? Five days to reach a million users.
Right?
Insane. Insane. How quickly this has been adopted. And you are right. It's interesting, the waves of technology that seem to totally disrupt businesses over the last, you know, ten or fifteen years have all come from the consumer side.
A hundred percent. A hundred percent. I mean, think about think about the march, of ChatGPT itself. Right? A hundred million users in two months, and it took Facebook what, Paul? Do you remember? Like, four and a half years to get there?
Yeah, and we thought that was a game changer.
Yeah. Yeah. No. Seriously. So, I think, I think generative AI is here to stay, and it it's thrust artificial intelligence, into prominence.
You know, I think I've shared this with you. One of my favorite LIMRA statistics, research, is around underwriting. Right? And the use of AI in the underwriting space.
In 2017, when we first started tracking AI for automated and activated underwriting, it was about 50%. Less than 50% of carriers had some level of AI in their underwriting programs. Right? We refreshed that benchmark in 2021, published in 2022, and that number jumped to 93%, Paul. So the AI revolution was ongoing within our industry, and then generative AI just thrust it into a common lexicon.
Yeah.
I also think, at the end of the day, you know, fundamental business economics will drive adoption.
Yeah.
You know, who wants to hire more underwriters to review more cases? Kartik, would you like to hire more, or would you like to keep the same group you have and have them make them more efficient? And I think when a company has to make a decision like that, the decision is clear. But it doesn't happen you know, as you say, it's not a straight line. There’s a curve, and I think we're at a huge inflection point at this moment.
I agree with that, Paul. So one thing I will say about AI and generative AI. Right? And I've said this to, to executives across our industry. AI/generative AI is not a play for hiring less people or reducing headcount. Right? It is a productivity play. Right? Imagine what you can do with 10% of your day back. Right? 20%, 30%, I mean, the promise for generative promise of generative AI is just that, to be able to give you, time back into your day. And if there are real repeatable operational things that you do as part of your job, it is more than likely that those will be automated by virtue of AI, right, in the next several years. So, I think our industry stands on the on the cusp, of significant transformation as do we as a species, thanks to AI.
So, Paul, here's a question for you. Right? We’ve talked about AI; we've talked about generative AI over the past several months. Let's come to today. What role is AI currently playing, across the insurance landscape?
Well, it clearly has, I think, won over the heads — of the minds — of a lot of people in this business. Hearts? I'm not so sure. I think we've got attention, which is good.
I think the level of which the industry has embraced this, I think it not only varies by company, it varies within companies. You know, there's certain units that I think are jumping on this stuff. I think marketing, you know, is something you can immediately use in a marketing area and do some just incredible stuff. I think some of the other areas there are — first of all, I think this will touch every single area of our industry.
Yeah.
I'm not sure how fast — it will go. You know, is it one year, two, or three years? Guarantee it will impact everyone everywhere.
But I think we're really in sort of that, storming, norming phase of talking about transformation. You'd agree?
Oh, a hundred percent. Right? And I think that also depends on carriers. I will say, I mean, we have the, LIMRA and LOMA AI Governance Group. Right? So, we started this group, what Paul, back in late December, early January. We have now up to 72 business and technology senior executives that we meet on a monthly basis. Last Friday of each month, Paul.
And we're trying to figure out what good looks like for our industry. Right? What are the best practices that we can bring to bear? Co-create for our industry. What are the use cases in play? What are the use cases that are actually seeing value? How do you measure value? All of these are big things for us to solve.
What's most interesting to me in terms of AI adoption, no matter how progressive organizations are in their digital practices, right, because everybody's been doing digital transformations, no matter how progressive organizations are in their technology stacks, no matter how many legacy systems you are, writ large, every single carrier has doubled down on vibrant use cases around AI. And I think that's remarkable. So, when I talked about hearts and minds, Paul, maybe the minds of everybody, certainly my heart. Right? Because I do see the significant value proposition that is going to emerge from this across the value chain.
Yeah. It's interesting. Like, you talked about the consumer use of ChatGPT. I found that people close to doing the work are in love with the possibilities. This is interesting. The best use cases have come from people who are literally on the front line, front line of sales, front line of taking phone calls, front lines of doing financial analysis.
I think what will set companies apart, as they implement it, is what's happening at the CEO level, what's happening at the C-Suite level. Yeah. Because to really implement this stuff and put it into production, you’ve got to make a lot of changes — changes in how your business strategy, thinking about your HR, how you're training people, thinking about how you're actually deploying some of these things. I mean, you know, the rules of deploying an application for ten years are not any different than putting one of these things into play.
Yeah. Hundred percent. See, you know, AI — you and I talked about this before — AI is just a tool. Right? It's a technology. It's a really sophisticated tool and technology. Right? But at the end of the day, it's a business strategy enabler, and not a replacement of anything. Right? So how you implement that is going to be the key. And, Paul, we've seen this through the AI Governance Group. I think what generative AI has done; is it has allowed CIOs to have a 360 perspective of all AI activities within their firms.
Right? So, as you know, Paul — I'm preaching to the choir, but our industry does one thing better than most, which is operate in silos. Right? So, we've gone ahead and built up vibrant digital practices, digital AI practices within underwriting. Right? No one told finance. You know, then we've done some AI stuff in actuarial, but no one told marketing. And we've effectively established these AI systems, for the past many years, which generative AI has cured.
Right? It's allowed us to get a 360 perspective, and only then can you achieve economies of scope and scale. But to your point, this has to be championed from the C-Suite on down across an organization. Right?
Yeah. Well, yeah, you're absolutely right. I think the group you’ve pulled together is phenomenal. It's been great. I think you've got a very interesting cross section of disciplines on your policy. And, you know, I'll date myself. I remember when, you know, oh my gosh — the Internet comes around. And I remember sitting in the discussions of, ‘Do we really need a website?’ That was the first question. The second question was, ‘Well, does Kartik really need the web on his desk?’ You know? ‘Let let's do a CBA for that.’ I see a lot of, parallels between that time and this time. Like, Kartik, can you imagine being, Oh, yeah. My job, I'm in charge of the Internet at the company. Silly. Right?
Do you remember a job called Webmaster?
Yes.
Yeah. Right. Right. So, you know, there are echoes of the past, but it's different, you know, it's a different era, different technology, but, you know, humans are always going to be humans in how they do this.
So, I think — and this is for a setback — I think the NAIC was very good in saying, you know what? Look. Let's require all of these companies to have AI committees to as — to your point — this is something where you can't just let something sort of peck away at one side of the business without understanding the implications, you know, on another side of the business. So good news is we've got a committee together. The bad news is, oh my, committees are hard — they’re hard.
Yeah.
Right? And I find in our organization that, you know, it can be — and I don't want to cast aspersions on anybody in anyone at the company — but it it's also a good excuse to say, oh, the committee will solve this.
Right. You know? Oh, I'm in, you know, data security. They are going to handle it. They are going to come out with a “no, no, no, you.”
Yeah.
You know? You need to change your job. You need to go out and take these courses. You should take this time, this little interim. We got a little bit of a window before — we know the wave's coming. Now is the time to get ready for the storm. Take the courses, learn, try this stuff on your own. So, I think, it's great where everybody's doing it together. It's just, also be aware that, you know, sometimes committees, you know, it means nobody's accountable. Nobody can say yes. A lot of people will say no.
What is that John F. Kennedy quote, quote? It was, “The best time to fix the roof is when the sun is shining.” Right?
Exactly. Love it.
Right? We need to prepare for AI. Now I also think, Paul, there is going to be an inflection, a little bit of an inflection, right, between general AI, and just for those listening. Let's maybe take a second and let's, let's demystify what we mean by AI versus generative AI.
Right? So, AI is a broad, sprawling field of study as you know, and includes everything from machine learning to voice and text-to-speech and speech-to-text. Right? Generative AI is one manifestation, one type of an artificial intelligence, and it's allowed AI to be democratized, right, across our professional and our personal lives.
So, I do see an inflection point coming very soon where generative AI is such an integral part of all of our vendor products. Right? So, whether your firm uses AI, or it doesn't use AI, whether your firm builds AI, doesn't use AI, you will be an AI consumer. Right? If you have Microsoft Office 365 and you get Copilot 365. Right? And that's pretty amazing. Right? You can actually give it a few prompts and say create me a slide deck. Right? A PowerPoint deck. You know, like ten slides, and boom, boom, boom. Right? It's done.
You know, we imagine the product that would be left there. So, where I was going with that is you're going to end up using generative AI as just part of the normal course of operation like we think about the Internet today. Yeah.
Well, I'll tell you. We put together a seven point plan for 2024 four to — I would describe it as enable — sort of lay the groundwork for the company to get into AI. Point number five was to open up one of these chat tools to our entire organization. So good stuff happens. Get somebody to raise their hand. They say I'm working on this IT department. We're going to put it together. Great. I got the memo back. Okay. They are maybe you're going to turn a tool on, like, by December 25th.
Like, okay. Wait till our next meeting. Now before I even get to the next meeting to your point, I go and log in to Microsoft Copilot and somebody turned on Copilot.
Yeah.
So, I talk to people, and they say, well, this is not ChatGPT. And I said, yes. It is. It's already here, and you turned it on. Thank you very much. To your point, we can pretend we have control over committees and structures, but you're right Kartik, it's here. It's coming. We're not going to be able to turn this stuff off. We better know how to use it.
Yeah. Which is why I think Paul, most organizations have and continue focusing on two things. Right? Number one, I think establishing some rules of the road and governance is important.
Look. Lots of organizations, and I would say, you know, just I I'd hate to put a number on it, but maybe two thirds of organizations have blocked access to public GPT-type sites from their corporate networks. Why? Because you don't want somebody — with all good intent — right, accidentally copying and pasting your intellectual property or asking a question that divulges corporate secrets.
Right? That's probably a super inappropriate use of ChatGPT. So, whereas most organizations are also working on implementing private versions of public GPTs. Right?
So, most organizations have their, have their internal engines, stand up as well. But I think look. I think guardrails are going to be very important. Educating employees on those guardrails is very important.
Right?
Yes.
Educating your employees on that, making sure people are AI literate, are going to be — is going to be — a critical thing for firms going forward.
No. I agree. I think education is key. I mean, education on you know, we tend to especially if you leave the state — you know, no offense to listeners in the IT or HR department — we tend to say, here's what you can't do. But I think it's ‘Here's what you can do.’
Yeah. Yeah.
I think it's very important.
Yeah. By the way, here's what you can do. How do you, you know, how do you write a prompt? Yeah.
You know, I have a laugh, Kartik, on the webmaster job. You know, I remember the big salaries now is, you know, prompt engineer is the latest hotdog. But you know what? There actually is a lot of power in understanding how to create prompts.
You know? Some people don't know how to Google, you know, how to Google well.
I said this to somebody the other day. People still don't know how to Google.
Yeah. Yeah.
Right? So, yeah. So, I think the best way to derive maximum benefit from these tools is to understand how to interact with them in a safe and, and secure manner. Completely agree with you.
Yeah.
So, I think education is what I think if any organization were to spend if they said, listen. We're just going to spend a hundred dollars this year. What will you do? Education. Open up courses up for your employees so they can understand. And, boy, if you have another hundred dollars, give them access to a tool they can actually use for work.
Yeah. Makes sense. Hey. So, let's, let's pivot to today. Right? What do you think AI is doing currently, in insurance?
Like, what's, at large? Think about the value chain. I know you talked about marketing, and that is your prism. But if you go across the value chain.
Sure.
What themes, what divisions, what use cases, domains are going to be most transformed?
Yeah. I think you mentioned the return. I think we have a lot of people who say immediately, let's jump the CBA. Let's do a ROI. Very, very hard to do that today. And, you know, one of your members, I think, I wrote this down. Kartik, most of my good ideas come from other people. Right?
That's the value of LIMRA and LOMA.
Right? Yes.
Let's all talk about good ideas together.
Yeah. I won’t name the person in the company, but they said, “Listen, we bucketed these projects under does it make life better? You know, give people back five minutes of their you know, give them fifteen minutes back each day.
That would be really valuable. Bucket number two, does it drive sales? Yeah. That's okay. Yes.
We don't exactly know how much we think it will. And the third bucket is, does it really change our cost curve? And I think that's the hardest one to do. So, I'll kind of walk you through some of the stuff we've been trying, and we had we started out, Kartik, at the beginning of the year with three POCs, all in the marketing area. Why? Because I had control over it. I could be a sponsor and the person doing it. So, the committee was a little easier to get approval.
Yeah. And, and, also, it was all public information. Right? You were taking stuff that we had on our website, you know, generating.
We actually got that POC out to beta in a production environment. We've been getting some really interesting feedback, and we'll talk about that more later. People know that. I think some of the other areas, I think the next area where I think there's a lot of experimentation going on, I think there'll be payoffs is the service and ops area.
So we, for instance, we've been purchasing a lot of little blocks of business, from carriers. It's interesting. We got this niche of — listen — if you have some New York you know, if you got a block of business in New York, you don't want to deal with it, send it to us.
Yeah. These blocks are complicated. Like, you know, we just purchased one where, you know, it oh, sounds like one policy. No. No. No. No. Eighty different policy types sold by, you know, three, four different types of companies. How do you educate the call centers to be able to answer these questions?
And I'm telling you, we're getting this thing to rip through these policies and be able to answer questions quickly.
We've also got some promising, you know, statusing is an interesting problem. You know, where's the pizza, once agents start to put the business in here? And, you know, we've got a we we've got a very conversant chatbot that will tell you not only about your business, but are you important to our business? How important are you? How many times have you rung the bell for our company over the last couple years? And there's a lot we can do there. So, it's not only service, it's service to sales, part of the our product area. Well, interesting.
We have these sites that everybody looks for and it's kind of, you know, the price line of annuities and insurance, and, how did how did this company, actually how does this product actually work? Sometimes, you know, we all don't fit into the four or five boxes that some of these companies illustrate our products. Let me actually compare the real language of these policies to each other. So, we've actually been able to read through, you know, policy files to better understand, you know, the differences between the products we're doing.
The most interesting one, I would say, was our investment department. Okay. How do you optimize portfolios for very specific products. You know, the annuities we sell are all eight, nine, and ten-year duration products. Well, you want to match the assets with those.
Okay. You've got all the capital chart. Okay. Yeah. I can get a, I get this great rate with this, you know, private, you know, credit vehicle, but, oh, yeah. There's a big capital charge and there's risk. So, I've got to diversify my portfolio. Yeah. I've got to figure out and optimize yield across a certain period of time. Yeah. It's doing it. The stuff is doing it. It's doing it. We’ve gotten the elements to actually do it verbally. Yeah. We've actually had it write the code for us to optimize this stuff.
See, that's the other thing. Lots of companies are using, code assist tools.
But you know what, Paul? Let let's, let's park here for a minute. Right? Because, I think, I I'd like to tease out some of the more successful use cases that we've been going through.
But also, everything good requires some level of guardrails. Right? There are challenges that organizations need to know, with AI and generative AI implementations. I'd love to tease those out a part. Everybody talks about fairness and transparency with AI, so we should talk about that a little bit more. And perhaps the best way to do that is to is to come back here.
If you ask, I'm here.
Hundred percent. Let's do it, brother.
Okay.
Thanks for listening to LIMRA's insider insights podcast series. To hear future podcasts, subscribe at LIMRA.com/podcast.