[00:00] [music]

[00:07] [music]

[00:13] Paul Lucas: Hello everybody. Welcome to Insurance Business TV. We are here at Insurance Fest 2026 for one of our peer-to-peer interviews. Now I'm meeting with a man who certainly knows a few things about technology. That man is Naveen Dhar. He is senior director insurance industry advisor at Microsoft. So Naveen, everybody of course is talking about artificial intelligence. What would you say is perhaps the most meaningful way that AI or connected data is already changing underwriting or claims today versus what would you say is still hype?

[00:43] Naveen Dhar: A good question, Paul. So, first of all, thanks for having me. If you start looking at this from, you know, let's take claims and underwriting and all the documents that come in the content perspective, that's happening today, right? There are tools that can help extract data. You can make the underwriting process, claims process much easier. You look at the co-pilot aspect of it where you have large documents that need to be summarized or you want to check for compliance needs or you want to ask a question about history that's there and it can help you act like an agent for you.

[01:18] That exists today. The connected aspects of it where a claim that could be a $2 million claim, a water leak is detected and resolved, that happens today. You know, you look at all of the telematics that exists and all of the products that are coming out of that like.

[01:37] Now, that being said, there's still a lot of stuff that is still a hype that people talk about, but we're nowhere close to that. Case in point, one of those is this whole concept of a robotic underwriter for specialty lines and commercial. Now, I'm not going to talk about other lines of business, but specially for these.

[01:56] We are far away from that. I think the focus right now is that for complex risks judgment is still critical, and human judgment is still required. So, assistance for underwriters is going to happen, but that's still far away. The second aspect of this is the whole notion of take an LLM and point to data, and it'll create magic for you.

[02:23] That doesn't happen, right? So, data irrespective perspective data is still king. You still have to clean up your data and get the right data in place. So, that aspect because models have existed, and models are not the challenge. The challenge has always been data. So, if you believe they can just use LLMs to make it all become a magic, that's not going to happen.

[02:42] So, that's some indication of what's going on in the industry right now.

[02:45] Paul Lucas: Yeah, and of course you're Microsoft, so you have I would imagine quite a huge overview of many different industries. What do you think the insurance industry is still lagging behind on tech?

[02:56] Naveen Dhar: Interestingly enough, I keep having people tell me this about insurance industry lagging behind on tech. And you know, there's some aspects of it that I agree with, some we don't. So, data science, right? It's a new term, but in insurance we've been doing it for a long time.

[03:11] We have used actuaries to do risk management since age. I think what's the challenge, I think the area is that insurance is very lagging behind is on the data aspect. Probably the connectivity aspect. We have had siloed systems where we had legacy systems that have existed for a long time. The carrier system, the reinsurance system, the broker system, they're all separate and disconnected.

[03:35] There is no way for me to have a connected system where if there is something that is found out by the carrier, can I real time connect it to the reinsurer? So, that aspect is still something to be worked on. That being said, right, I think the bigger piece here really is about insurance.

[03:56] You know, we have had specialty lines where the volumes have not been high that people have avoided doing automations because of that stuff. We have had scenarios where you have a legacy book of business that is doing well and there was no need to change something that ain't broken, right? So, why change something?

[04:16] And then the tech skills, sometimes tech resources have chosen to go to the tech industry than come to insurance part of it. But having said that, I think there is change coming through. Especially with all these data science aspects which we have been really good at is something that insurance will continue to do well and you'll see a lot of stuff happening.

[04:36] Paul Lucas: Yeah. And of course, everybody wants to scale these tools as well. Is there any big barriers standing in our way from being able to do that?

[04:45] Naveen Dhar: And so, there is this whole concept of pilot purgatory, right? So, where people are creating multiple pilots because ultimately the biggest barrier is the business model around which is rightly so, which is we want to make it sure it's trustworthy, explainable, and so we're making the right documentation available.

[05:06] So, if it's audited for the regulators, we can show them how we came up with the decision. Because it can't be a black box that gives you a decision. You can't track why it made the decision. So, people have been doing a lot of these pilot regulatory pilots, but some companies continue doing many pilots and have very few things that have actually gone to production.

[05:31] Versus others who have not done as many pilots, but whatever they have done, they have tried to take a small portion over to production and they have learned how to manage all of the regulations and the trustworthy aspect of it and all the auditing part of it.

[05:48] And I think if we were to look at this and change the way we start thinking about it, I think that will help eliminate all the barriers we have. But the biggest barrier right now is for us to document and show whatever we do is trustworthy, has auditability, it can stand the test of time.

[06:06] Very rarely have I seen somebody stand up to the Department of Insurance and say AI told me so, that's why it's working, right? And for that to stand, they won't accept it. So, having that documentation, building that trustworthy stuff, building that traceability is going to be critical.

[06:25] Paul Lucas: Yeah, if you don't mind, I mean, you talked about AI. I think a lot of people used to see that as futuristic. It's very much here now, of course. But if you were to look ahead over the next 5 years, what does an insurance tech stack look like and perhaps some things that are not in our natural lives at the moment?

[06:41] Naveen Dhar: Yeah, so if you look at today, insurance today, right? Insurance today, the tech stack is more like a filing cabinet. Right? I take stuff, I put in there, I take it out and check it and put it back in. I think that whole concept will change. We will go away from a filing cabinet to a trusted colleague.

[06:59] And what does that mean? It means that I will be able to have an agentic layer on top of our core system. I'll have specialized agents for claims, for underwriting, for different parts of the business that I can work with. I will be able to get to a point where I can have a conversation.

[07:18] Like today, I have 200 fields to fill out. I could have a natural language conversation with AI and then those fields are being filled up. There's going to be new tech required. You know, look at AI liability, look at all of the neat things that are coming with cyber, you know, with these models being able to detect all your vulnerabilities and being able to patch your vulnerability.

[07:38] Cyber insurance is going to change. So, that's going to be a big piece. Climate, there's going to be new tech stack to be created for climate insurance. So, all of these aspects are definitely going to come through.

[07:52] Going back to the scale and the connectivity aspect also that's going to be, I think you asked that in the first question, which is and I think it's also going in the future. So, I may now have real-time risk model because today when I do risk, I take a snapshot annually and then come and look at a snapshot the next year at a renewal.

[08:12] But there's no way for me to do continual risk assessment at continual risk. So, if a sensor alert goes off that updates my risk model. And oh by the way, since it's all connected also sends a prevention message so that the prevention action is taken. Moving us away from incident and repair to prevent all incident, right?

[08:36] So, preventing how the claims can happen is going to be the key. So, this whole ecosystem connected between the carriers, the brokers, the reinsurers, getting all of these ecosystem connected ecosystem is also something that's going to happen.

[08:50] All in all, I would say new intelligence layer on top, natural language processing to talk to agents so they become more of a colleague than a filing box or a filing cabinet.

[09:03] New connected devices ecosystem that's real-time. Now, we do have connected devices, but there's no real-time transaction going on between different parties that will be a key. New insurance products that are coming out and will require new stack whether it is climate, cyber or AI liability risk.