conversational AI

DRUID Talks Ep #6 AI in Insurance – A Behind the Scenes Look with Ema Roloff

DRUID Talks Podcast Ep. 6 explores the current challenges and opportunities the insurance industry is facing and how AI technology helps address both.

Episode #6 of the DRUID Talks webcast features Ema Roloff, Insurance Expert, multiple award winner and social media influencer, and Subject Matter Expert Kieran Gilmurray. See the full episode and transcript below.

Subscribe for the next episode!

Kieran Gilmurray: Folks, welcome to DRUID Talks! In this new episode, we explore the current challenges and opportunities the insurance industry is facing and how AI technology is helping address both. We will take you on an exploration journey through industry-specific use cases to discover how AI technology is utilized, improving underwriting and actuarial science, looking at customer experience, combating fraud, calculating risk, contributing to increased operational efficiency, as well as improving customer and employee experience.

Our guest today will share with you some great advice and some lessons learned from a very successful career implementing AI and insurance solutions in multiple organizations. We'll also look into the future trends and share thoughts on how you can best leverage the power of AI in your insurance business. Welcome, Emma Roloff, Insurance Expert, multiple award winner and social media influencer. Delighted to have you on board today!

Ema Roloff: Well, thank you for having me, Kieran, and I think you might have caught me doing a little bit of a dance to that groove there before I thought I was back on camera. But I'm glad to be here!

Kieran Gilmurray: Well, here we are. Good to have you. Let's jump into the very first question. The insurance industry is known to be risk-averse, or at least it has a reputation as slow-moving. Has COVID or present-day AI possibilities pushed insurers to move forward with AI and all sorts of digital technology?

Ema Roloff: So first, I would say a little bit of that impression of insurance being slower moving than other industries, I think, is a bit of an unfair rep. So, you mentioned that I've got experience working with insurance across a couple of different companies within my career, and prior to my current role, I worked in a cross-vertical role where I was working with manufacturers, companies that were focused on biotech and kind of in the medical space, as well as insurance. And I think the deep, dark, dark secret that nobody wants you to know is nobody's actually as far along in the adoption of technology as we would like to pretend or portray based on the front-end processes that people are seeing. So it's one thing that I always share with the insurance industry is that we're probably doing ourselves a disservice by pretending we're really far behind instead of focusing on where we've seen adoption and how we can increase that adoption.

So it's kind of like my first disclaimer. But I would say just the same as any of those other industries, COVID really accelerated the adoption of technology across the board and in insurance specifically, when we start looking at things like AI, doing things like visual inspections and leveraging technology where we could when that person-to-person connection was no longer an option, is definitely something that we've seen.

And there's a number of companies that kind of like we're in a different industry that made their way in. So, you know, especially with things like that example of property inspections, people that were operating in the construction arena all of a sudden leaping over to insurance because there was a need and a quicker adoption cycle for this type of technology to be able to come in and do these video inspections and have it be immediately accessible, whereas that barrier of entry into the industry would have probably been much, much larger had COVID not happened.

Kieran Gilmurray: Well, let me ask you a second question now that we've dispelled the myth that insurers are behind every other organization on the planet. All organizations experience challenges and opportunities when it does come to adopting and then particularly scaling AI within their operations or building new business models. What are insurers' biggest challenges and biggest opportunities?

Ema Roloff: So, I would say one of them, and it probably presents an opportunity, but it's starting with a challenge. And a lot of the conversations that I'm having today are centred around the looming talent gap that is emerging within insurance. So, for those that aren't aware - anybody that lives in the insurance industry probably has a pretty good idea of what I'm talking about - but it's estimated that over the next five years, there's going to be about 500,000 jobs available with insurance that we don't necessarily have the talent to help fill. So I think it equates to about 15% of the workforce retiring in the next five years because, going back to a little bit of our reputation issue within the insurance industry, it's not the most exciting and flashy in terms of bringing in younger talent because they might not be as aware of what opportunities are or how they can use skills that maybe they see as like technology or fancy, exciting things that could translate into insurance.

So that conversation of the talent gap - there's a lot of companies that are asking the question of how do we use technology to augment the people that we have? So that this talent gap isn't as wide for us to fill. So, I think that that is a big driver behind bringing in AI and just automation technology in general to kind of help fill that need. But we also recognize that we can't completely absorb that with technology alone. So, there's, you know, there's some work beyond that from a challenge perspective.

I would say the other big area of opportunity that everybody is having conversations about within the industry is how do we start to really, truly, leverage all of the data that's coming from connected devices and how we… I mean, insurance has so much data… When you look back at the longevity of policies in certain scenarios. I was doing an interview on my show the other day talking to someone in the life space, and they said that they had just talked to a customer about their longest-standing policy; they have a 103-year-old who has a life insurance policy with them, and the policy dates back to when they were in their twenties.

And think about the amount of data that is associated with that policy potentially over that life span and access to all of the rest of this data. We don't have the infrastructure to necessarily bring it in, make sense of it, and drive decisions that are going to really, truly change our path and how we go about things like new product development and that kind of thing without doing some groundwork to get us there.

And so I think a lot of people are asking the question of, like, okay, so how do we make sense of all of this data? How do we leverage it, and how does that change our product development? How do we shift from just responding to risk to being able to predict it and prevent it? That's another big part of this through this additional data coming from all of these IoT devices. And just like the AI models that we're able to produce today, there's a lot of that conversation of how do we make that flip from risk mitigation to prevent and protect truly with all the data that we've got.

Kieran Gilmurray: I don't know where we're getting more data from. Is it the 103-year-old or an IoT device? So you're sort of wondering.

Ema Roloff: Depends on what kind of data you're looking for.

Kieran Gilmurray: Say, we’re all thinking IoT… Well, I personally am delighted about the old story, and it is very tongue in cheek, and I did work in insurance for 12-13 years and loved every moment of it. I'm delighted to hear there’re half a million jobs open. There's a chance for me yet to go back. And then it's not a case of, you know, banks or insurance companies where good people go to for their careers go to die. So, it's good. But give me an example where insurers are embracing AI, and if they are embracing AI with this rich data, that you were… the rich data that they have available, what are they doing? Or when you mentioned earlier on, you know, regulatory things might inhibit how insurers go about things… are they embracing or are they reluctant?

Ema Roloff: So, I think most carriers are embracing technology and the viewpoint I think of… might have to take a step back. So, one of the big things that I've seen within the insurance industry, probably more prevalent than in other industries, is this mentality of building their own solutions and building their own IT infrastructure. So, you know, it's not uncommon to find insurance companies that are still running on legacy platforms that were literally developed before I was born or when I was a small child.

And the… and I mean, again, going back to… there are plenty of other industries… the Southwest incident is a perfect example of the fact that that is not unique to insurance. But this mentality of building, maintaining, and advancing their own proprietary technology is something that's really close to the central culture of the insurance industry as a whole. And I think that mentality came along with the initial adoption of AI, and there are many insurers and specifically large insurers, that have their own data science teams, and they're building their own models, and that's part of what they see as their proprietary advantage as a carrier. And that's not to say that that's not a great model if you've got the right talent in-house and that you are able to do that effectively.

But one of the things that I'm starting to see more - and probably since COVID - is a willingness to adopt solutions that have been developed and bought and brought into the organization as a way to accelerate the adoption of AI. So, it's not that there weren't these programs put in place and that there weren't people looking at AI and machine learning as a part of their underwriting structures or how… really kind of across the board.

But we've seen this acceleration when this mindset of build vs buy is starting to shift. And even going back to core applications, there is a mindset of starting to shift towards that idea of buying software and bringing it in and having, you know, your secret sauce is how you use that technology and how you pull everything together. So, I think that that's helping with the acceleration a lot.

And where I've seen some of the biggest use cases… So I mentioned things like property inspections, you know, back… pre-COVID, the idea of drones taking away the claim adjusters job entirely was a big discussion. But there are many companies that are starting to leverage drones and computer vision to really be able to build these impressive predictive models or modelling structures, being able to use them in the case of a catastrophe to go and do an initial survey. Maybe when a physical adjuster can't get there if it's not safe in those types of scenarios, we're seeing high adoption… and then it is becoming an option more and more for things like visual inspections of your home if something happens or like a car after an accident during your first notice of loss being able to attach images or videos of what's happening and having total loss predictions done against those images or estimates in terms of what type of damage was created and helping you figure out where you need to get your car fixed.

So those kinds of physical limitations that we had before in the industry are one of the first areas that I've seen high adoption. Then other big places for things, as I mentioned, like total loss prediction, but things like fraud detection and prediction or litigation prediction are big areas where - especially on the claims side, we're starting to see more and more adoption of AI - because those resources that are spent using their gut intuition and their experience in the industry to comb over all of the information that's coming in as a part of a claim - are either ageing out of the workforce or they're really expensive resources. And when you are constantly, you know, there is a red flag, we're going to send this one to take a peek at. Those are expensive resources that are highly trained. And if it actually didn't have a high likelihood of fraud, you're wasting that time and money delaying your customer from getting paid what they owe and really detrimentally impacting your customer experience from that standpoint.

So even though it seems like something that's meant for back-end processing and saving the carrier money, it actually is to the advantage of the policyholder because if they're not fraudulent, they're getting paid quicker, and they're not having to jump through as many hoops to get a claim settled because we're able to use technology to say, “Oh yeah, this is one that we owe them a payment for. We're going to get it processed and get it processed quickly because we know that it's got a low-front score.” So those are just a couple of things that I've seen.

Kieran Gilmurray: I like that. I like the claims one because I used to use the phrase, “Working in insurance is a kind of distress purchase.” In other words, I'm distressed when I have to buy it, but I know I need to. So I can imagine all of that translates into better customer experience because I would want my claims settled pretty quickly. And if the cost of actually managing an insurance claim is reduced, then potentially, we share a little bit of that benefit. The insurer makes a little bit more money, and the customer policy is able to be priced a little bit more competitively.

Ema Roloff: And when we look at customer experience too, I think it's something like… I saw an Accenture stat that I think it's only 29% of people report being happy with their current insurance provider. And a big part of it is just what you said. It's a need. It's not a want for a lot of people. Nobody rejoices after they buy a new car for having to go get a policy to cover that car. That's something that we aren’t excited to purchase, but we know that we need it. And then the moment of truth is really that claims process. So you hope that you buy it and you don't ever have to interact with them. You hope that there's never a case where you have to go and proactively engage in a customer experience with your carrier, at least on a B and C… well, and especially on the life side.

But when you start to look at how quickly people will switch providers if they've had a poor customer experience with you… think it's like 46%. So at least 50% of people say that customer experience is their number one buying factor when it comes to an insurance policy. And if you mess up that claims process, the likelihood of them renewing with you the next time… they're going to jump like that.

And so anything that you can do to improve that customer experience, especially on the claims side, make it as easy as possible for them to buy so you can reduce any heartburn with this necessary purchase and then take as good of care of them as you can on the back end to really come forward and, you know, put forward your promise and show them that that promise was worth all the premiums that they've paid over the years. If you can't do that, then you're not going to have customer retention. And in this environment that we have with like a competitive landscape, you're just not going to be successful.

Kieran Gilmurray: I’m thinking about that 50% based on my one and only claims experience in the last number of years… You mentioned a moment ago, if you don't mind asking the next question, that you know there are instances now where insurers have started to use technology that they might not have before. So, for example, using drones to get to inaccessible areas or to make quick decisions on AI. Where are you seeing inside of insurance companies, Ema, potential uses for AI in insurance where maybe they're not actually being touched by those insurance companies yet, but they very well might?

Ema Roloff: I think there are a lot of areas where AI has potential within the industry. I would say one of the probably, and I wouldn't say it's not touched because there are people… but this goes back to that idea of really mining all of the data that we have from connected devices. Almost every carrier that I talk to is talking about building out kind of pay-as-you-go or not pay-as-you-go, but like pay-how-you-drive or, you know, really looking at giving you discounts for connected devices in your house and how they can start to bring in all of this data and information that we're gathering and really change the way that we're managing product development. And so every carrier is playing around with this idea. There are a few who have kind of figured it out, or there's, you know, like, there's the notoriety or the notoriety that comes with like Lemonade or Metromile being acquired by Lemonade. These examples are really AI-driven companies. But when you start to look at the combined ratios of some of these companies, there are people within that industry that are pretty sceptical of the longevity associated with some of these; in short, tech startups and then traditional carriers are all playing around with these ideas, and they may have discounts or loyalty programs and that kind of things that are there. But I don't have a sense that anybody has really figured out the perfect balance of how to make that work. And I think we're going to see much more of that in the next couple of years as some of these pieces all start to come together.

I shared an example, and I TikTok-ed a couple of weeks ago… My husband, James, and I just replaced our Nest doorbell with a Ring doorbell and a security system in our house. And our insurance company had this big campaign that you get a discount if you use a Ring. And it had to be a Ring because they have some sort of affiliate relationship. So, we didn't ever qualify for the discount when we had our Nest doorbell, but we switched. So, we went through the process of, you know, we're preventing the risk that's, you know, at least you're not getting access to our Ring doorbell information. But at least we have a security system. We're preventing risk within our homeowner policy. They gave us an annual savings of $1. That's it! And so, again, and that's not enough for us to necessarily switch providers but think of the… like our time is worth more than that $1 on an annual basis. And I can't imagine how much you spent on that campaign to try and get us… And now that's going to sit in the back of my head. So, that's one strike. And again, if something were to happen and we were to file a claim, and that claim wasn't managed right… that right there, I don't need a 3-strike, I'm out, and we're going to go find someone different. So again, we're still trying to figure out how we reward this behaviour that's mitigating risk. How do we take in this telematics information?

You and I did a “Third Thursday” episode a couple of months ago where I said I don't necessarily want a connected device in my car that's going to show you that I speed. So like, yeah, you can give me a discount for having this here and being able to see what my driving behaviour is. But, like, why would I go and allow that if it's potentially going to be detrimental to me as a policyholder? So we're still, I think, have to work through what that reward structure is going to look like and how we really, truly integrate this data into our product development in a way that's beneficial to the customer.

Kieran Gilmurray: Remember, I mean, we are being recorded, so you may want to reverse backwards in case industry carriers are listening to this…

Ema Roloff: I’m not that terrible…

Kieran Gilmurray: Potentially speeding, folks, that it wasn't an actual speeding; it was just a theoretical.

Ema Roloff: But I mean… Bryan Falchuk, who is on the call with us, also mentioned this idea of, like, you know… so if I hard brake to avoid an accident, it's measuring that kind of information in my car, and I avoided the accident... they're not ever going to know that I got into that accident. But maybe I live in a city where hard braking is a regular occurrence because you're getting cut off on a regular basis. So, I mean, you know, I'm avoiding, I’m being a good defensive driver and avoiding an accident, but it's going to make me look like a poor driver just based on that data alone. Now, you could make the argument that someone who lives in a city who's getting cut off on a regular basis should have that as a part of their policy. And it's probably already in some of the, you know, how we're calculating those policies in the back-end of traditional models. But at the same time, from a policyholder perspective, that's not a motivating factor for me.

Kieran Gilmurray: It doesn't sound like it to me. Well, let me ask you the next question, then, just out loud. What would your advice be for insurers who want to implement AI or automation? Because if I'm an underwriter and I'm an actuary, or I'm a company that’s building this... it's the volume of data and the potential that's huge. But then, because it's so big, it can become overwhelming as well to know where to begin, where to focus, and what not to do. So what would you be advising insurers to do or not to do?

Ema Roloff: I think the first message, and Kieran, you're probably going to smile because you've seen me go through this… like screaming from the rooftops for years across the board. But anywhere that you're looking to truly, like, build a scalable technology, whether it's, you know, traditional automation or AI anywhere along that spectrum, you need to make sure that your basics are covered.

So again, going back to within the insurance industry, you're looking at, if you have legacy technology that's at the centre of everything you do, you're going to encounter challenges when you're starting to bring in massive amounts of data and newer technology. Just even from an integration perspective, it's going to provide barriers. But again, going back to some of the basics, what does your data model look like within your organization? Do you have a data warehouse set up? Have you gone through a transformation process in that capacity? And I think big carriers certainly have, but mid-tier to smaller carriers, that might still be something that needs to be addressed. And so really making sure that you've got some of the technology foundations laid down before you jump right to AI. And I understand that people are being told that they need to go directly in that direction. And that's not to say that you sit on the sidelines until that's done, but you do need to make sure that you have the foundation in place.

And then the other piece that I would say is important when you're building out this strategy is to have true visibility into where all of the trendlines of your technology footprint lie. Because as one of the challenges that are emerging - now that carriers have started to get more comfortable with the idea of buying solutions and bringing them in-house - it's only exasperating silos, and really like this idea of all of these point solutions existing and claims, being able to go out and buy something for themselves and the product team having different tools that they're using and quickly, you can lose sight of real value that can come from scaling your program if you don't have it centrally managed in some capacity or at least have a strong understanding of where tools are being leveraged and how they're being leveraged across your entire company or across the ecosystem in general of insurance. And so that's that that idea of point solution kind of overload is something that is another big conversation that's being had within insurance. And where does the burden of integration and data sharing lie? Is that on the carrier? Is it the job of the insure-tech companies to figure out how to all play nicely and build these, you know, these big ecosystems solution sets where a carrier can go, okay, I'm going to grab this?

And so we're starting to see that come up now that people are more comfortable with going and buying. So it's, again, really making sure as an organization that you have a strong foundation, that you understand where you're bringing in technology across the entire company and where you have the ability to scale it and leverage it in multiple areas.

And then probably the last piece, and this is just good practice in general from a business perspective, is having a really clear vision of where you want to take the organization so that when those decisions are potentially being made in silos, they're at least all headed in the same direction. So, you have at least everybody moving in generally the same area so that you're not trying to pull everything together when you finally do get your act together to figure out how to scale things in all of these different directions. So that vision, I think, is incredibly important from a leadership perspective as well.

Kieran Gilmurray: Fantastic! One last question, if you don't mind. We hear a lot about ChatGPT and generative AI, and Bard and other solutions are available. What can ChatGPT or generative AI do for insurance?

Ema Roloff: I've seen this conversation popping up a lot, and I actually heard a carrier talking about their own internal policies of when and how it's going to be allowed. I think that there are big possibilities on the sales side of things in terms of communication back and forth with customers. I think the biggest challenge that's going to come about, and we've hinted at it a couple of different times, is what do regulators say about all of this and how does that because there's a lot of hesitancy on the insurance side to jump into new technology without clear guidance from regulatory bodies. Because especially in the United States, it's not federal regulation necessarily; it's state by state. So even if you're leveraging something in one state, you might not be able to in another. And it just starts to open up this big can of worms for people, and you know, ChatGPT is really great at aggregating information. Still, unless it's integrated with all of your internal policies and information that you've got, you might not be providing accurate information when you're using it in the form that it is today.

And so, you know, I think the big opportunity is when organizations start to bring it in-house and have it using our own language and our own information to help potentially present the right information to a customer. I've also seen questions, again, from the regulatory standpoint, but also from carrying risk, like, “Are insurance companies going to care or insure inaccurate information being provided by AI? And how does that start to really like complicate things even further?” But within the insurance industry, we're still trying to figure out if we can communicate with policyholders in a binding way through text messages. There are certain states that won't allow that kind of thing. There are certain states that still require physical signatures on stuff. So even being able to automate something as simple as a signature is a challenge!

So then, when we start getting into these, like true emerging technology use cases, I think there's a lot of hesitancy to jump with both feet. And that doesn't mean that there aren't people evaluating it. It doesn't mean that there are not tons and tons of use cases that we'll see pop up over time. But I think it's an area that a lot of carriers are kind of tiptoeing around while they figure out where it lands exactly.

Kieran Gilmurray: I wonder if our 100-year-old life policyholder noticed any difference in the insurance industry, or in the industry, for that matter, over the last 87 years.

Ema Roloff: I mean, just think about what that person has seen over the course of their life. It's like the Queen of England when you start to see all of the things that she witnessed in her lifespan. It's insane to think about what their perspective on transformation may be.

Kieran Gilmurray: Well, hopefully, you and I lived 103. We may have different perspectives along the way. Emma, thank you so much for coming to DRUID Talks today. If people want to find out a little bit more about you or follow along to get some expert advice, how did they go about doing that?

Ema Roloff: So, you can find me on LinkedIn. My LinkedIn URL is like that kind of standard one with just where I am pointing here. My name is here. So it only has one “M”, which is the tricky part. And then, like I said, I do a lot of TikToks, so you can find me on TikTok. You can search “Ema Digital Transformation”, and I should pop up, and I have a YouTube channel. But all of those links and stuff, if you make your way to my LinkedIn, I have a get in contact with me. You can hit that, and you'll see all of my social as well as be able to reach out to me directly.

Kieran Gilmurray: Well, folks, you have a choice, which is the best thing. So like, thank you, everybody, for coming to DRUID Talks. Thank you, Emma, for talking so expertly about the insurance industry is making a big splash in insurance. It's excellent to hear that. I think insurance is as far ahead and potentially as far behind every other organization that there is on the planet.

It's great to hear some of the use cases of the claims, the customer experience, the drones, and everything else. It's great to hear that there are half a million jobs potentially coming in the insurance industry as well. That's really exciting. It's good. I have to say that I love this industry. I have been in it for years. I think it's very underappreciated in many ways. I can't wait to see what insure-techs, insurance companies, or a combination of both actually do. But the future is bright for insurance, and I look forward to seeing it at this age and maybe as a 103-year-old policyholder in the future.

Thank you very much, everyone!