Paul Carroll
With AI transforming underwriting, some say the function is entering a new phase. Do you agree?
Katie Klutts Wysor
When you look at what property and casualty carriers are saying and what brokers are starting to say, there’s broad recognition that generative AI could reshape risk assessment and underwriting in meaningful ways. It’s becoming an enabler. It can help underwriters make better decisions and work more effectively.
Carriers are focusing on underwriting as a function and investing heavily in it. They’re talking about that focus in investor days and earnings calls, and they’re doing significant work internally to organize data and update processes—improving speed, increasing automation, accelerating turnaround times and supporting more informed decisions with better data.
What’s even more significant than the process improvements is what AI could do more broadly. Think of the underwriter as managing capital and trying to direct it to the ideal place. How can AI help define the parameters around what the carrier wants to write so it can deploy capital where it is targeting stronger returns? From there, you can align appetite and business mix with the underwriting process.
For example, if I’m a regional workers’ comp player and I want to expand into other lines of business or other states, how can I use AI to support that portfolio management decision and direct capital more effectively? Then, how do I identify the necessary distribution partners to find the business I want? How do I create the proper incentives for distribution partners to bring forward that business, so I have a submission to underwrite? And then, how do I make sure the underwriting process, and decision-making aligns with my appetite?
I think that’s where a great deal of value could be created from an underwriting perspective, looking at how AI can help inform research on the front end, and how you then use something like a GenAI-enabled underwriting platform to begin systematically embedding strategic capital decisions into appetite, process and guidelines at scale, so underwriters evaluating risks are working from more relevant information.
Then you can use AI to respond to new business decisions more quickly, respond to renewal decisions more effectively, and potentially take certain actions during a policy’s term to support risk mitigation conversations.
If you can start mastering that link—how you’re deploying capital and setting appetite, all the way down to those micro process decisions—that represents a new level of maturity.
Paul Carroll
Speed has become a significant competitive factor in insurance underwriting. If you’re slow to quote, even with a slightly better price, you may lose the business. What’s happening in terms of speed in the underwriting process, and how do processes need to change—not just the technology—to take advantage of the speed AI can offer?
Katie Klutts Wysor
Speed to quote and bind means something very different across varying lines of business. In auto insurance, you need to be able to deliver a quote in seconds. So, you see many personal lines players focusing on quote simplification. In auto, it is close to a mature problem, and many carriers are following established market patterns to stay competitive.
But the speed question gets harder as you move into more specialized or complex lines in personal insurance and small commercial, middle market and large commercial.
What we’re seeing there is impressive. Capabilities are now emerging to triage submission intake. From a technology perspective, carriers increasingly can take a submission no matter how it comes in—via email, a platform or another channel—and combine it with what they already know about that risk, along with relevant third-party data.
At that point, it becomes an execution challenge. How can you more systematically get to a quick yes, no, or maybe on appetite, and then move effectively toward a quote?
How fast that can happen depends on the line of business. But that point is right: Carriers should move as quickly as their competitors. If you’re slow, distribution may not be willing to wait.
And the benchmark will continue to move.
Paul Carroll
How far along is the insurance industry in using technology to allocate capital more intelligently, and what needs to happen to reach the next level?
Katie Klutts Wysor
It’s less a technology challenge and more a business decision-making challenge. Some players in the market are especially strong at this, and you can see that by looking at underwriting returns over time. Those companies have consistently used technology and data to manage their portfolios and allocate capital with greater precision, and they will likely continue to adopt new approaches as the tech capability improves. The timeline for broader adoption of newer technology, including generative AI, is harder to predict because it comes back to how quickly "the pack" of carriers can evolve how they manage the profit and loss across the portfolios.
Paul Carroll
What is an example of how insurers can improve their capital allocation?
Katie Klutts Wysor
The fundamental approach is to look at your underwriting returns against the capital you’re deploying to the business, map that out, compare outcomes, and decide where you want to grow and where you may need to pull back. Improvements in technology may allow carriers to do that analysis more frequently. Instead of doing it once a year as part of strategic planning, you could be looking at a refreshed view every month using more current data.
Many carriers may be able to move from annual reviews toward monthly or weekly review cycles, depending on how they make decisions. They may also be able to do the analysis in a more automated way and make decisions more intentionally on micro-segments of the business (by geography, class, line, etc.) that would have been too time-consuming to identify and react to previously.
Maybe a competitor enters the restaurant space aggressively and undercuts on price. You may decide not to follow them down that path because you believe the pricing is unattractive, so you slow growth in restaurants.
Or take the opposite scenario: A trend affects restaurants and causes the market to become more cautious. You may conclude that the market reaction has gone too far and decide if this is the right time to pursue restaurant business.
Today's leadership reviews may only look at class code-level data monthly or quarterly, and at frequency and severity trends in a backward-looking way. But if you can automate how, you assess that information at a portfolio level, then you can decide whether to lean in or lean out of a class like restaurants more quickly.
Paul Carroll
What about the execution side of this—how do insurers actually act on these faster insights once they’ve identified an opportunity?
Katie Klutts Wysor
The second half of the equation is exactly that. Say you’ve been able to automate and generate more timely underwriting data, so you can make portfolio decisions weekly or monthly instead of quarterly or annually. That’s a meaningful shift. The next step is execution.
Say you decided to lean into restaurants. You want the market to know. You want your agents and distribution partners to know you’re interested in that business, particularly if another carrier has started to pull back or take rates. That’s the business you want to enter the pipeline.
Then you want to set up your underwriting process so you can pivot quickly. Maybe you were not prioritizing that business to get it to an underwriter’s desk and streamline escalation paths to support faster turnaround.
Of course, once submissions are flowing in, and the process is in place to evaluate and price the business you want competitively, you also need the proper governance and controls, so you don’t end up writing risks that fall outside appetite.
The big difference this year versus a year ago is the ability to put agentic AI workflows in place and that support faster transaction-level decisions. Humans are still in the loop, but they are not necessarily slowing down the process in the same way they did when carriers relied more heavily on manual referral and escalation processes to respond to market changes. The next frontier I expect to see in the coming year is using agentic AI workflows to help improve portfolio-level decisions.
Paul Carroll
Would you talk a bit about some of the process efficiencies from generative AI as underwriters make their decisions? While those efficiencies aren’t as strategic as the portfolio-level decisions you’ve described, they still seem substantial.
Katie Klutts Wysor
Underwriters face a series of yes, no, and maybe decisions, and much of the friction sits in the maybes. You can automate obvious yes-or-no decisions. The maybes are the gray area where you bring in a person.
Over time, we may be able to bring in a person less often because of agentic AI and other decision-support tools, while maintaining appropriate human oversight.
Some maybes exist simply because a piece of information is missing. A file gets routed to an underwriter to obtain one additional data point. Once that data point is available, a rule can be applied, and the case can become a yes or a no. In many cases, that is increasingly solvable today.
You should identify those cases in your portfolio, then use AI to obtain the data point and apply the rule.
There are also maybes that are more judgment-based, where you’ve created a manual review because you want someone to look at it who has seen this kind of case many times before. Maybe they’ve seen a six-figure loss in a similar situation, so you ask, “Would you write this again knowing what you know now?”
Agentic AI workflows can help by bringing more context to the situation and supporting more informed underwriting judgment.
Paul Carroll
Based on what you’re seeing, how much are underwriters working with brokers and clients to provide guidance on risk reduction—essentially telling them, “you’re getting dinged for this, why don’t you fix it to reduce your risk?”
Katie Klutts Wysor
Right now, it’s predominantly brokers and distribution partners that are providing that first line of risk management advice. But there’s also a meaningful role for underwriters and carriers.
The concept is there, the question is how consistently it can be translated into actionable guidance.
Paul Carroll
What final advice would you offer to readers?
Katie Klutts Wysor
What my clients care about is taking some of the bigger ideas and translating them into what to do right now and how to respond in practical terms. So, what I’d leave readers with is this: Keep thinking about the art of the possible but also focus on what you can implement now to strengthen performance this year, and start bringing those two together.
Understand what technology, data, and AI capabilities are available to you. But more importantly, identify which ones you can deploy quickly while you continue building toward the more complex architecture and data challenges you should address over time.
Paul Carroll
So you can create a cycle: make targeted investments that create near-term efficiencies, then use those gains to support the next wave of investment.
Katie Klutts Wysor
Exactly. Don’t spend three years trying to build the perfect solution. There’s a lot you can do right now. Deploy something practical that can create value, then use those gains to support larger investments over time.
Paul Carroll
Thanks, Katie.
