While our coverage at ITL these days can feel like "all AI, all the time," there's still a massive open question: How quickly can we develop and adopt AI agents that can act on our behalf?
Most of the articles submitted to me are from enthusiasts, but there are reasons to be cautious, too. We've all seen or read about the hallucinations AI can have, about how AIs can learn to be ugly, and so on. How do we know they won't do something stupid or offensive if we cut them loose? One nasty lawsuit or loss of a major customer can outweigh a lot of efficiency gains.
Insurance will get there with AI agents. The question is just how fast. If you could tell me today what adoption would look like at the end of the year, I could tell you a lot about the state of innovation in insurance. So I'll be watching closely.
Six other areas also seem to me to be key for this year. The two other huge ones are the spread of the IoT, which is letting us monitor so many more issues in real time, and the growing adoption of the Predict & Prevent model, which helps turn that new data into ways of protecting customers. Embedded insurance and parametric insurance are also playing increasingly important roles — and have lots of potential still. And autonomous vehicles have built a launching pad that could let them take off this year, with all sorts of implications for insurers. Finally, I'll focus on the general notion of speed. Sometimes, a quantitative change is important enough to make a qualitative difference, and I think some processes in insurance, such as underwriting, may start moving so much faster that they could change the competitive landscape.
Let's have a look at the seven keys I'll be scrutinizing this year — and think you should be watching, too.
AI Agents
My caution stems from the sort of problem I've witnessed in years and decades past related to "decision rights." Remember when "Internet refrigerators" were supposed to track what you had in them and order, say, milk for you when you were running low? Well, I didn't want my refrigerator ordering for me. Maybe warn me if I'm running low on something or be able to tell me when I'm at the store whether the thyme is still usable, but that's it. I control what I eat.
I'd love for AI agents to take repetitive work off my plate, but I give up control slowly.
My optimism stems from the sort of vision that my old friend and colleague John Sviokla described in a webinar conversation we had recently on AI. He described AI agents as employees that we'll be able to supervise and monitor as we do our current staff and suggested that each of us might have dozens of highly trained AIs whose autonomy is carefully circumscribed. He said job interviews in the future will include the question, "What AI agents do you have? Show me your bots." If you don't have an impressive array, John said, it will be like interviewing for a chef position at a fancy restaurant without having your own set of knives.
AI agents will be a big one to watch.
IoT
The Internet of Things, especially if you include auto telematics, has already made a massive impact on insurance. Drivers are being coached to be safer. Devices are detecting fire hazards so well that carriers are giving them away to customers. Water leak detectors in homes are getting closer to that tipping point, too, where carriers will give them away because they'll prevent so much damage.
A sort of operating system for homes now lets any sort of device communicate wirelessly with it and have a signal relayed, letting you know about that water leak or that your smoke detector has gone off while you aren't home. Amazon's Sidewalk creates what's known as a mesh network that allows wireless connections to all sorts of nodes, even in public spaces, that can relay a signal to wherever it needs to go. Meanwhile, sensors just keep getting smaller — I wrote in November about how researchers had even managed to outfit Monarch butterflies with trackers weighing .06 gram.
Progress will only accelerate.
Predict & Prevent
When I got involved with Insurance Thought Leadership a dozen years ago, following decades of writing on innovation and technology, the first talk I gave carried the title, "He Who Sells the Least Insurance Wins." My reasoning: Nobody wants to buy insurance, but everybody wants safety. So why focus on selling insurance when the industry can take its massive amounts of data and expertise and provide safety?
That's obviously easier said than done, but I've been delighted to see that there's been so much progress and that the industry is rallying around the Predict & Prevent idea (sometimes called by slightly different names). At ITL — and more broadly at The Institutes, of which ITL is an affiliate — we've tried to highlight some key examples, such as Whisker Labs' Ting, which detects electrical faults in homes and is preventing hundreds of fires a year, and Nauto, whose two-way cameras in truck fleets are drastically reducing accidents.
In 2026, we'll highlight as many more as we can — while hoping for more breakthroughs.
Embedded Insurance
Embedded insurance had a big year in 2025, to the point that toward the end of the year authors published two major articles with us on the topic. One said embedded insurance was nearing a tipping point and marshaled an impressive amount of evidence about the companies leading the way. Another said embedded insurance wasn't just a way of reducing distribution costs but had become a key part of the customer experience, by simplifying the purchasing process.
There are some complications. For one, agents will resist being cut out of the purchasing process. For another, carriers that offer insurance as part of the purchase of something else risk ceding control of the experience to the seller of that product. The insurance could even be treated as a commodity, meaning one carrier could easily be swapped out for another.
But the convenience is still so great and the drop in customer acquisition costs so substantial that I expect more and more insurance to become embedded.
Parametric Insurance
Parametric insurance is showing up, in particular, in agriculture and in natural catastrophes, where it's relatively straightforward to find an agreed-upon metric, such as wind speed or lack of rainfall, and where a speedy, partial payout on damages can be key to keeping a business running or to rebuilding quickly.
We haven't covered it as much as we might have, but I suspect we'll see parametric insurance make inroads in lots of areas in 2026.
Autonomous Vehicles
AVs have had their ups and downs in the nearly 13 years since Chunka Mui and I wrote a book on the topic, but they seem to finally be on a glide path — and an exponential curve at that. While Uber, Cruise and others have fallen by the wayside, Google's Waymo keeps expanding relentlessly. It's up to 450,000 fully autonomous paid rides per week in the U.S. and expects to hit 1 million a week late this year. (Waymo tends to understate its goals and routinely exceeds them, unlike, say, Tesla, where Elon Musk has been promising full autonomy for a decade but only has perhaps 30 robotaxis on the road at the moment, almost always with safety drivers.) Waymo keeps expanding into more cities and will even move into London this year, where it will go head-to-head with China's Baidu.
Amazon's Zoox has begun offering limited service, as have some startups, including May Mobility and Nuro. Nvidia just announced big plans to supply the technology and even much of the AI for car makers that want to develop AVs, debuting with a slick offering it developed with Mercedes. Tesla, of course, continues to talk big — and much of Musk's potentially $1 trillion pay package depends on meeting aggressive plans for AVs.
AVs have taken hold enough that an ER doctor — as in, not a techno-optimist — recently wrote an op-ed in the New York Times arguing that we have to move as fast as possible to AVs for public safety reasons. He argued that AVs are now so clearly safer than human drivers that we have no choice.
I think it'll be a big year for AVs, whose effects will eventually trickle down not just into auto insurance but health, life and workers' comp.
Speed
One of the impediments to innovation in insurance has always been that it takes so long to see if an idea will pan out. In Silicon Valley, when they talk about A-B testing, they're talking about testing thousands of different ideas — headlines, offers, etc. — every second. In insurance, if you want to test a new price or new product, regulatory approval alone can take months, no matter how fast you go internally.
But generative AI has increased the metabolism in insurance, and I think we're just beginning to pick up speed. Because Gen AI can gather so much information so fast and at least pre-process it, claims agents and underwriters can make decisions faster than ever before. Carriers can also start experimenting with automating certain classes of submissions and claims so that a human never even has to touch them. Agents can respond to customers faster, too, and speed everything along.
Those who increase their speed the most will win a competitive advantage, according to all sorts of surveys showing how much customers value quick payments and how carriers and MGAs may lose business if they're slower than the other guy at responding to a submission.
Increased speed could cause even more fundamental shifts. For instance, here is a piece we published recently on how underwriting could move so fast in some cases that no binder would be needed. Just think about how much work that could eliminate. Or, consider what happens if policies aren't just reviewed at renewal, and underwriting becomes continuous, updating as circumstances change. There are all sorts of implications, as Bobby Touran and I discussed in a recent webinar that carried the confident title, "Continuous Underwriting Changes Everything."
Speed is such a powerful but broad force that it's hard to see just where it takes us, but I'm sure it'll be somewhere interesting and important.
Here's to a fascinating 2026!
Cheers,
Paul