← 返回

2026.10.09 | TB | Insurance | Travelers: AI & Technology Strategy

原标题：The Travelers Companies – AI Opportunity & Technology Strategy

#financials

Third Bridge · 2026-10-09

Analyst-led • 6 Oct 2026

Companies: TRV US, PGR US, LMND US

Bharat Sri Harsha Karpurapu (BSHK) Principal AI Architect, AI Transformation at ServiceNow Inc

Jacob Zuller (JZ) — Third Bridge Sector Analyst

Agenda

Key Insights

Introduction

JZ [00:00:03]: Welcome to Third Bridge's Interview entitled The Travelers Companies – AI Opportunity & Technology Strategy. I am Jacob Zuller, and I'll be hosting today's Interview. I'm here with Harsha Karpurapu, former Director of Generative AI and Insurtech at the Travelers Company.

Harsha, before we get started today, please state I agree or I disagree to the following statement: You understand the definition of material non-public information and agree not to disclose any such information, or any other information which is confidential, during this Interview.

BSHK [00:00:34]: I agree.

JZ [00:00:35]: Would you mind beginning with a brief introduction to your background?

BSHK [00:00:40]: Yes, sounds good. As Jacob just mentioned, I was the Director of Insurtech Partnerships and AI Innovation at Travelers. Part of my responsibility involved looking at AI innovation across areas like underwriting, claims, search, and even the multi-agent systems, and most importantly, yes, generative AI. My primary background is primarily in building and scaling the applied AI systems in large enterprises, particularly in regulated environments. Of course, I have worked across traditional machine learning, NLP, computer vision, gen AI, and more recently, I am also one of the major architects of agentic AI in a big SaaS company here.

Questions

00:01:25] How are you seeing the insurance market right now? How is AI changing the industry and insurance carriers down the road?

BSHK [00:01:42]: Yes, definitely from the insurance market overall, I do see that the market is softening a bit. Of course, I would not say it's uniform because property has clearly moved into much more competitive environment, while casualty, particularly US casualty and commercial auto, remains much more disciplined. In a hard market, carriers have the pricing power. However, right now, as I feel like we are entering majorly into softer markets where carriers need to compete on the risk selection, speed, distribution experience, and the overall efficiency. All in all, without destroying the underwriting discipline. Of course, if you look at it, the way AI came in as an advantageous tool, particularly for major insurance carriers in a softer market like what we have right now.

00:02:42] Where do you see AI being a differentiator in the softening insurance market? What shifts could we see in a hard market?

BSHK [00:02:55]: Yes, definitely. For example, if you look into the AI differentiation, I primarily look into four areas where it can differentiate. You can start off with submissions, as soon as the submissions has been coming in. How do you map it to an external data? How do you summarise the risk? How do you identify the missing information? Probably how do you triage the information and recommend appetite along with routing the submission appropriately, so that carriers can respond materially faster without necessarily losing the underwriting standards. I don't think the differentiator is not about who has the chatbot because when people look at AI, majority of the times, they talk about, "Hey, we have this beautiful such solution, or we have this chatbot that can answer your questions instantly," I would say the major differentiator will be coming into, how do you integrate this AI into the operating workflows, and combine it with proprietary data, underwriting expertise, and also the underlying decision systems.

Also, if I have to summarise it, the differentiation, I see in four areas, risk selection, underwriting speed, distribution experience, and expense efficiency, of course. When we look into the harder markets, I would say the technology priorities are largely the same, but the economic emphasis shifts. Because in a hard market, carriers can only… I would say carriers mostly focus on risk selection, because the harder market is also the seller's market. They focus mostly on risk selection, portfolio optimisation, and maximising the attractive capacity. However, when you look into a softer market, speed, operating efficiency, the broker experience, and making sure that we are giving the best experience for our customers without spoiling the underwriting expertise, it's all much more important in the softer market there.

00:05:24] You talked about submissions and summarising the risk. How do you see AI improving there? How was it pre-generative AI vs what could it look like down the road?

BSHK [00:05:46]: If you look into pre-gen AI. Again, when we talk about the pre-gen AI and post-gen AI, I want to also mention that AI has been there. It's just that the naming has been different. What ChatGPT has done was they put a web interface on top of this intelligent model and democratise that access to every person. This way, people with or without AI knowledge can able to access that intelligence through the chat. That's where I feel like, because of the OpenAI guys, that the AI or the generative AI came into the existence or came into the picture. If you take a step back before AI, we have deep learning. We still have the language models. Of course, we did not call them as large language models. We have the BERT systems. We also have T5s from Google. We have these language models has been existing there. However, what this AI, the latest gen AI models have given is a multistep reasoning approach. It gave the humongous intelligence into the hands of these carriers. Because earlier, we were using the language models as individual for, "Hey, summarise this document. Look into this image and let me know whether it's a single-family dwelling apartment or whether it's a commercial unit."

What this latest intelligence models have given us is an ability to look into the output and orchestrate the next steps after looking into the output. You can imagine that before gen AI, I don't want to say before AI, I would say before the latest multi-reasoning AI, we have to do all these approaches manually. With the advent of this multi-reasoning AI came into the picture, we are actually accelerating the existing workflows with the reasoning mechanisms. Earlier, it might be an analyst or an operating person looking into the loss runs manually. Of course, we use some extraction mechanisms with the language models available at that time. Now, with the latest intelligence model, we are having language models looking into these loss runs, extract this information and, of course, query external sources to populate the missing information. If you look into all these things, AI is not the solution for the ultimate outcome. AI-powered workflows. AI is acting as an acceleration to get to a particular output. That's where I would say the before and after experience.

00:08:49] How are you thinking about The Travelers Companies' position when it comes to AI, Innovation 2.0 and its position down the road?

BSHK [00:09:03]: Again, we all know that because Travelers have also won this latest claim CIO competitions with their claim AI solution. We all know that Travelers definitely moved beyond the experimentation stage. In fact, very recently, they also released their own insurance LLM. Definitely, there has been a lot of work that has been done in AI inside Travelers. If you look into Innovation 1.0, it was largely about building the organisation capacity around data analytics, digital process, and modern technology. When you look into the 2.0, it's about using that foundation to apply AI across the insurance value chain. Very recently, Travelers also publicly mentioned that almost 20,000 of their employees regularly use AI. There has been millions of transactions are automated, and agentic AI is already embedded in their operations. I would describe Innovation 2.0 as moving from individual AI use cases towards AI being part of the operating model. It's also looking at how Travelers' employees are changing the ways how they use it to do the work.

00:10:33] You brought up that Travelers developed its proprietary LLM [large language model]. How do you view that strategy vs partnerships with leading LLMs? Is there room for both? How do you think about those dynamics coming into play?

BSHK [00:11:01]: That's a very good question. Increasingly, not only Travelers, for example, even the big SaaS companies and other tech companies also just signed a letter with US government for open weights. You can see that the industry as a whole, yes, we have these frontier tech models available. They are also increasingly interested more to develop their own language models. A couple of things. The one thing is if you have your own language model, you don't need to be vendor lock-in. That is one thing. Also the privacy needs. Increasingly, there has been cases about the data breaches. There has been some lawsuits. Probably, particularly in regulated industries, not only Travelers, I might see some banks or even other insurance companies, they might be developing their own language models. When I say they might be developing their language models, those language models might be having the intelligence that is enough to cater the insurance sector or that is enough to cater their internal use cases.

However, the moment when they need multistep reasoning or multiprocessing orchestration, they might go with a hybrid approach. When it's more like when something has to be reasoned across multiple steps, they might use the frontier tech. When something is as easy as extracting an information, summarising a document, they might revert those requests to their internal proprietary LLMs. This way, they get the best of both approaches. Privacy whenever it's needed and the super intelligence reasoning approaches whenever they want, they can get by approaching the frontier tech. I would see the industry will be moving around a hybrid approach, in short, just marrying themselves to some frontier tech there.

00:13:13] Where are you seeing the biggest efficiencies in claims and automating that because of AI?

BSHK [00:13:26]: Particularly, when you talk about the claims, I would start off with… Before I get into the claims also, just looking into the near-term efficiencies. As I was saying, Travelers also started with by rolling out their AI tools across 20,000 employees. Clearly, we could see that they are trying to change the ways how employees use it to work. Probably the near-term efficiencies are around intake summarisation, document handling, even information retrieval, triaging, routing. That's where I see the near-term efficiencies. If you look into the claims, particularly claims, it has humongous amount of unstructured information. We have notes, documents, images, and sometimes videos, calls, and even the policy language, historical claim information, there is a lot of unknowns that might involve when you look into a claim. Particularly, gen AI is very effective at turning that information into something that an adjuster can immediately act upon. As I was telling you earlier, carriers are not using the AI to make a decision.

Carriers are increasingly using AI to crowdsource all this information. Fill in this missing information, provide the final summarised version and adjuster so that the adjuster can take that decision. The final step, ultimately, it's going with the human-in-the-loop approach. Ultimately, the pace at which a human can make the decision is what accelerated through the AI, particularly in the claims. The biggest opportunity is eventually orchestrating several of these steps together rather than automating each one independently there. I do know Travelers also… It's already publicly available information. They also launched agentic AI Claim Assist and AI claim analysis tools, which are mostly towards summarising these claim documents and some internal knowledge assistance.

00:15:45] Is there a way to contextualise how much claims could be either a straight-through process or fully automated? Where are we currently, and how long could it take to get that last bit automated?

BSHK [00:16:03]: For that one, most probably, I would resist putting 1% across the whole claims book because the automation ceiling varies dramatically by claim type, and also the technology is improving… It's running at a much faster pace. You can see us the low-touch claims. Simple, low-severity, high-frequency claims with clear coverage and liability have much higher potential for straight-through processing. As soon as you introduce bodily injury or litigation, fraud indicators, and some complex commercial exposures, human judgement becomes much more important. I expect a significant increase in straight-through processing only for the low-touch claims, which are low-severity and high-frequency ones, but probably an even larger increase in AI-assisted claims, the workflows what we were referring earlier, where AI crowdsources and slices, dices all the information. We will see more increase in those AI-assisted claims workflows where the adjuster remains accountable for the final decision.

00:17:29] What are your thoughts on how AI could improve, and how Travelers could see savings on the loss adjustment expense ratio or underwriting expense ratio, and how do you see AI improving those down the road?

BSHK [00:17:59]: If you talk about how AI is improving or changing just savings there. Definitely, I wouldn't say the biggest economic benefit from LLMs is frequently not removing an adjuster or something. It is removing the administrative work surrounding an adjuster. Thinking about a 200-page claim file, summarising prior notes and stuff like that. AI can encompass a lot of that work. Also, when you look into if AI can compress a lot of the work, and we are looking at more number of submissions a person can underwrite. We will see majority of the top line increase while also decreasing the expenses for the bottom line there.

00:18:57] How do you view AI as a catalyst to underwrite more policies? What impact do you see from an efficiency perspective in risk selection and pricing?

BSHK [00:19:23]: I strongly feel underwriting is probably one of the most strategically important AI applications in the commercial insurance. Because historically, a significant portion of the underwriters' time just consumed through gathering this information, rather than making these underwriting decisions. In fact, I would see AI start by the way, if you look into the underwriting workflows, you can look at it changes the way how we extract this information by submissions and also augmenting this with external data. Even once the data is there, it can also map it to the internal risk guidelines to see whether this particular submission is in the part of the appetite or not. Ultimately, it's going through, we will have a lot more underwriting capacities and greater consistency. In the longer term, I see the biggest opportunity is the risk selection. If you can combine this proprietary loss experience, submission information, and external data more effectively, you can differentiate risks within a class instead of pricing primarily from broad categories.

Also, one more thing I personally feel is, imagine if you have the AI looking into your risk guidelines internally or your risk appetite guidelines, and you majorly see that most of the policies which we declined belong to a particular risk class. That also gives an opportunity for the carriers to see, "Okay, we are having majority of these policies from a particular… Sorry, we are losing majority of the submissions because we are not writing particular class." It's about giving the carriers an opportunity to revisit their risk appetite guidelines and see how they can capture some of the missed submissions or the missed policies.

00:21:36] How do you see the opportunity to leverage AI for better risk selection and identifying missed underwriting opportunities panning out for all carriers? Do you see Travelers in a better position? As all carriers invest in, leverage AI and depend on LLMs, do you see all carriers selecting the same risk because it's better for each company?

BSHK [00:22:08]: That's a very good important distinction. It also goes with what AI tools these insurance carriers are using internally. For example, historically, we know that Progressive has been exceptional at this data-driven pricing and automation. Chubb's competitive strength is deeply tied into underwriting expertise. AIG has also been very public about this AI-assisted underwriting. What we just spoke about is identifying the missed opportunities. I would say majority of the carriers, looking at the pace at which the industry is growing, they might be already looking into these missed opportunities. Because at the end of the day, AI gives a lot more underwriting capacity because it's giving back the time to the underwriters to write and look into more policies. With that additional time, carriers will look into how do they increase their top line. I see that it will become a level playing field.

00:23:18] How do you view potential pushback from a regulatory perspective? How much of a risk do you see on the AI underwriting side for Travelers?

BSHK [00:23:40]: The regulatory pushback, that's actually a very good point. Because not only from the insurance regulatory bodies, we will have… As most of these AI solutions go into the mainstream, particularly in the regulated industries, they need to have a way to explain the AI solution side, the explainable AI. I would say that carriers can expect a moderate amount of pushback. Increasingly, they will invest in a lot more tools that can provide that transparency and regulatory documents that is where it can explain AI decisions much more accurately. I would say that's where you will see majority of the investments, more of explainable and traceable AI decisions.

00:24:39] How much of a factor do you see data quality, legacy systems and those parts impacting Travelers and its ability to leverage AI?

BSHK [00:25:03]: That's actually a great point, because before few years, people increasingly care about model quality. Over the past few years, model quality isn't the limiting constraint anymore. We know that the limiting constraint is whether the model has reliable access to the right data and that can safely interact with the systems where work actually happens. I feel you will see a lot more emphasis on the data quality and ability of the models to integrate with the legacy systems. Imagine this, if the customer policy, claim, and underwriting information sits across multiple platforms with inconsistent definitions, putting a sophisticated layer on top definitely don't solve the underlying problem. That's why you see some AI demonstrations look great on the marketing brochures. The moment we try to put these AI solutions to solve an underlying problem, that's where you see a lot of problems, be it in terms of the data quality or be it in terms of integrating with the legacy systems or, to your point earlier, how do we monitor and trace these integrations and these AI decisions? I would say it will become a lot more important one in this AI operating era.

00:26:42] How do you view AI impacting retention? In the past six quarters, Travelers' domestic business insurance retention rate has remained stable at around 85-86%, and you have the soft market where retention is usually a bit stronger. Do you see retention being driven higher because of AI, or is it more of a soft market factor? What shifts could we see?

BSHK [00:27:35]: I strongly feel retention will be higher with the AI coming into the mainstream. Just think of this way. With these new latest technologies, if you can identify and if you can predict a customer risk action. For example, with AI, I know some of the carriers are already monitoring the digital footprints of customers. Instead of before they go for the renewal, waiting for the policy documents for the customer to come in, as they're already monitoring the digital footprints of the customer, they are able to provide the right policy or the right premium that can be more enticing for the customers. I would say, increasingly, AI will play a huge role in retaining these customers because it's ultimately providing the better experience and better overall efficiencies for the customer.

00:28:44] How long do you see AI taking to retain customers? Do you see it taking a short amount of time for AI to capitalise on that opportunity or taking a little longer? How do you think about it from a timeline perspective?

BSHK [00:29:01]: From a timeline standpoint, it's about how quickly carriers are able to adopt these AI-powered solutions. Also, if I look at how the industry is changing, there has been a lot of FOMO about the AI solution adoption, not only across the insurance sector, like, every major sector. Already carriers are increasingly adopting the AI-powered solutions in the retention workflows. I would say that it's already there. It's already been acting upon. What I would see the major difference is, for example, Travelers built their own Travel LM [TravelersLLM]. Those workflows will become much more hybrid. Earlier, it might be just dependent on the frontier LLMs. It might be replaced with their in-house LLMs. That's where I see a majority of the shift that is happening from the tech standpoint, but in terms of providing better experience for the customers for increasing the retention, I feel it's already there, and it's just like we will see a lot more happening in the next few years.

00:30:17] Is there a way to contextualise how much improvement in retention AI could have? Do you see it driving it to 90%? Do you see it being just a small improvement or a large improvement in retention? How do you see it potentially down the road?

BSHK [00:31:21]: Definitely, I was looking at the Travelers Q2 results, even I moved out there. The results are pretty good, but again, at the end of the day, I don't think I could provide an exact estimate on how much the retention number is going to increase, but one thing I could guarantee is as carriers will get a lot more AI-powered workflows to understand the customer journey and the customer behaviour, you will see probably the customer hopping might also be decreased because existing carriers itself, they will provide a lot more better offers to retain the customer. Because in general, we all know that acquiring a customer is harder than retaining the customer. If carriers can provide the better services to retain the customer, which will be becoming much more attractive with the AI, you will see that those retention numbers will keep on increasing.

JZ [00:32:27]: You see AI improving retention more in a hard market than a soft market, like you brought up, shopping around where maybe retention could be a little bit better in a hard market?

BSHK [00:32:44]: Yes, yes. Retention would definitely be better in the hard market.

00:32:53] How do you see Travelers' position when it comes to having the talent to capitalise on AI vs peers, and how much of a factor is that?

BSHK [00:33:16]: To be honest, to start off with… When you say talent to capitalise, are you talking about the employee talent?

JZ [00:33:27]: Yes, employee talent, whether it's employee talent or even leveraging partnerships. How do you see potential strategies of the next generation of AI?

BSHK [00:33:42]: Sure, sure, definitely. Increasingly, you see that Travelers has been well-positioned with this AI-driven investment. If I'm not wrong, 2025, they also mentioned about the USD 1.5bn or something into AI and other technology initiatives in 2025. Yes, if I'm not wrong, that's one. Yes, it's USD 1.5bn. Also, you see that they rolled out AI-powered tools to 20,000 employees. I strongly feel like Travelers, there is the fact that change starts by changing the ways how they used to do the work. I see that they are upskilling their employees a lot by having them access the AI tools and also investing heavily with those numbers, like USD 1.5bn into AI and other technology initiatives and recently with the launch of their own proprietary insurance LLM itself. We could see that clearly, they are doing a lot for their employee empowerment and even developing their in-house LLMs and all these things.

00:34:58] How do you see AI enabling premium growth for Travelers? How do you see AI helping agents, brokers and the company for growth down the road?

BSHK [00:35:20]: Your question is mostly how do you see AI provide the better broker experience and things like that?

JZ [00:35:29]: Yes. Maybe how that could lead to premium growth. I know we talked about some enabling better risk selection, but how do you see it from a traditional growth perspective?

BSHK [00:35:47]: Yes, yes. Again, I can talk more from the AI side. AI can definitely contribute to the growth indirectly by making Travelers easier to do business with. We can see independent agents, brokers frequently submit the businesses to carriers. What you can expect is the response time will be drastically reduced. There will be a lot more appetite clarity, ease of quoting and the amount of information required can influence where business ultimately goes. If AI can reduce the submission friction and allow a carrier to respond faster while maintaining underwriting discipline, I feel that can improve significantly the hit rate and also the broker experience, ultimately giving a lot more business to Travelers.

JZ [00:36:45]: Do you see that growth accelerating because of the response time being reduced, or do you see it maintaining the same growth path?

BSHK [00:37:08]: I would say not only with Travelers, but the industry as a whole, I would say the response time is actually decreasing. For example, let's think about the high-volume commercial lines, especially cutting your process from hours to minutes, it can significantly change the carrier opportunity to get that business. Because ultimately, even if you look at what AI is doing, crowdsourcing (ph 37.38) this information, slicing and dicing this information, it's ultimately decreasing the response time what a carrier can reach back to the broker. I would say it will significantly increase the amount of business for any carrier, including Travelers, if they can have the right AI-powered workflows.

00:38:11] How do you see AI risks? How much of an opportunity is there for AI insurance? Do you see a need for AI insurance as a product? How much of an opportunity is there to insure potential AI damages? Currently, a lot of them are being excluded from traditional policies.

BSHK [00:39:06]: That's a great question. Yes, I do see that as an opportunity. Increasingly, we might not be surprised tomorrow, there might be an insurance for AI agents. We never know because currently, we already have the cyber insurance and it seems like Travelers already realised the importance of the cyber insurance on the cyberspace. They acquired one of the cyber insurance companies very recently. I don't have the name on top of my head, but yes. I do see this as an opportunity where a lot of carriers will be looking at.

JZ [00:39:47]: Would you expect Travelers to lean in there? To note the name of the company, it was Corvus Insurance Holdings…

BSHK [00:40:01]: Yes, yes, Corvus Insurance, you're right.

00:40:05] Would you expect Travelers to double down and capitalise on the cyber insurance opportunity, or how do you see a potential opportunity for the company there?

BSHK [00:40:20]: I would expect them to look into the opportunities adjacently because if you look into the cyber insurance, the way I look at this AI insurance or insurance for AI agents, it's an adjacent market opportunity to the cyber insurance. If you ask whether Travelers would introduce a product in that lines, probably they might look at any of the start-ups or the companies that might be playing the route, that might be trying to sell the insurance and look how they are playing for a couple of months and then take a step. That's where I would look at it because we know the AI agents, yes, they are becoming increasingly intelligent and also the number of cases with AI going rogue and the lack of governance around AI. It's also making it probably a hard ask for insurance companies to insure an AI agent. I would say they will be looking at that opportunity because it's adjacent to cyber insurance. Most importantly, they might be monitoring the start-ups, which might be operating in that zone.

00:41:44] How do you see AI shifting distribution? How much of a risk is that for Travelers and how consumers get insured?

BSHK [00:42:23]: Definitely, you see that AI is already changing how the distribution works traditionally, be it in terms of reducing the friction between the carriers and the agents or the brokers. If you look at the way it classifies the submission, identifying the appetite or even pre-filling the data or even think of one more user experience, as soon as for the small commercial customers, as soon as they start to fill in these details, automatically, the AI on the fly recommends the appropriate products. Imagine an Amazon experience, of course, I'm not talking about national accounts and stuff like that, just for the small commercial customers. I see that AI will change the way how the traditional distribution works. We will see a lot more changes in the user experience, be it in terms of pre-filling the data, be it in terms of recommending the products or prioritising the opportunities. Ultimately, it will go into shortening the quoting timelines. As I was telling, you will see a lot more in the commercial lines.

Over time, I expect AI to make distribution more personalised, even by helping the carriers to identify which risks, agents and segments are more attractive. It goes back into identifying the missing appetite opportunities, or identifying missing classes where carriers can do a lot more business if they could figure out a way how to write a risk class. Again, being said that, I don't think AI will replace the broker relationship as a whole when we talk about the AI-powered user experience and stuff because brokers and agent experience is going to be ultimately heart of any of these carriers, especially given national accounts or any complex commercial insurance. What I would see is it makes the distribution experience much more productive by reducing the administrative work and giving both sides faster access to the better information.

00:44:47] Do you see brokers and agents getting disrupted by AI more on the personal insurance side?

BSHK [00:45:00]: More on the personal insurance side. That's a great point. Probably brokers and agents, yes, will get… I would say not a some, probably some good amount of competition through these AI-powered user experiences. I would see that, but however, in a commercial P&C, I don't think it will happen because we know a lot of the business is distributed through independent agents and brokers. However, for the personal insurance, I do see that it will happen eventually.

00:45:42] Do you view the disruption of brokers and agents by AI as a risk to Travelers as we envision a future for personal insurance? If a consumer uses ChatGPT or Claude and searches, "I want the best auto policy," and puts in their information natively in that platform, what does that world look like? How does Travelers play there?

BSHK [00:46:18]: That's a great question. Increasingly, we know the smaller, I would say, insurtech start-ups are coming in. One of the start-ups are like Lemonade. It's literally like go to the… Open your mobile app, fill in the details, you get your insurance. Similarly, you are seeing people are even trying to shop for Amazon products and all these things through ChatGPT, Claude and stuff like that. Right now, ChatGPT has nearly 1.5 billion users. It's increasingly becoming an internet gateway for a lot of the users. Also, if you look into the demographic, increasingly, you are seeing the younger dynamic, I would say the later on millennials and the early gen Zs who are becoming the important customers for these carriers are using these AI tools on the day-to-day basis. Ultimately, I do see Travelers or any other major carriers will need to figure out to get to the customers where they are.

The customers are increasingly shopping around policies with ChatGPT and Claude or even Perplexity like that. Yes, they need to get to the customers and they need to figure it out. Then talking about Travelers as a whole. Travelers is a major player in the commercial P&C. Yes, personal insurance is there, and it's also not the big portion of their business. Their ultimate backbone is the business insurance. I strongly feel Travelers is very well safeguarded there. Also the people when I look into who buys Travelers policies, they are actually a different risk class than a Progressive customer. I would increasingly see players like Progressive, AAA and carriers like that will get a lot more fierce competition than Travelers.

00:48:49] What does Travelers need to do in the AI-native setting? Does it require an MCP [model context protocol] or another connector where a consumer goes to Claude or ChatGPT and asks for a policy and all the carriers with that connector pop up? How does that potentially work?

BSHK [00:49:23]: Only Travelers might be part of that universal insurance MCP connector. I wouldn't be surprised if OpenAI or Claude will go to major insurance carriers and say, "Hey, now almost everyone who want to shop for insurance policies are coming to our platform, or even Gemini. We are already having MCP connectors with other carriers. As we have MCP connectors with other carriers, we are able to show only policies from them. This is why you are losing like so-and-so policies." If you go to Travelers shop to start off with, they might say, "Hey, we have regulations and stuff." The moment if these big frontier tech are able to attract the smaller insurtech companies, the Lemonade and other companies like that, you will see that slowly the bigger players will come on board where they expose their quotes also through an MCP connection. Again, being said that, the technology is changing every day right now. The regulations around how this AI should work is also changing almost on everyday basis. To summarise this, yes, there will be a potential opportunity where other major carriers might expose their quotes through an MCP, but I would not say it's an opportunity which I would look at in the next couple of months. I would highly doubt that, but again, the industry might prove me wrong.

00:51:26] How much disruption do you see from companies such as Lemonade to Travelers? Do you see a larger incumbent carrier in a better or worse position to capitalise on AI vs a smaller, newer tech-native firm?

BSHK [00:51:54]: The way I would put it, again, having a start-up environment, you can travel much faster. You can embed much more AI features because your overhead is less and your decision fatigue is less. Being said that, as per my experience and even looking at their Innovation 1.0 and 2.0 policies, Travelers does have their own innovation teams internally. Pretty much they made sure that they put substantial technology investments. Also, you can clearly see that they launched a lot of claim-related workflows and even underwriting submission-related workflows. Yes, smaller companies like Lemonade might travel faster, but the bigger companies like Travelers, even though they might come up with a few features, but the impact of those features is humongous, where it can offset these smaller player changes.

00:53:11] What do you view as the top risks for Travelers down the road?

BSHK [00:53:23]: I would say the underwriting discipline, as parts of the market softened. Because we can see that down the line, probably after a couple of years, every carrier might have access to the same intelligent models, might be building AI-powered workflows. Think of this way, when the capacity is plentiful, the temptation across the industry is to pursue the growth, by giving back to the (ph 53.54) much price. You see that the industry, it's about how do you maintain that underwriting discipline as the market continues to get softened. Also the third one is, to your point, the risk execution around AI. The opportunity is definitely enormous, but moving from dozens of useful tools to truly redesigned end-to-end workflow, is much harder than building these individual applications. Because at the same time, the FOMO is real. It's about how do you navigate that FOMO by creating these intelligent workflows that are truly compliant, governed and give you the ability to make those informed decisions.

00:54:47] What's your overall outlook for Travelers over the next 1-2 years?

BSHK [00:54:54]: We could see that from the latest announcements, Travelers does invest substantial amount into technology. It's about looking at how can they get into this measurable operating leverage. One thing is Travelers does have the scale, even the proprietary data, even the domain expertise. If it can connect those capabilities into end-to-end workflows, especially around underwriting, claims and distribution, the upside is not simply on the expense savings as we were talking earlier, it is about the better risk selection, faster decisions and potentially more profitable growth. The next few years, I would focus less on how many AI applications a carrier has and more on whether these AI applications are visibly changing the cycle time, productivity, the loss outcomes and even the customer or the distribution experience.

JZ [00:56:05]: Is there anything else worth noting about Travelers?

BSHK [00:56:11]: We pretty much summarised a lot about the carrier experience and insurance industry as a whole, but one thing I would say is right now, AI is increasingly commoditised. Now it's about whichever carrier can design these outcomes-powered intelligent workflows will have an edge.

JZ [00:56:42]: Great. With that, we can end the Interview. Let me close by saying thanks again for taking the time, and have a good rest of your day.

BSHK [00:56:51]: Thank you. All right. Bye-bye.

Transcription ends at 00:56:56 of the recorded material