Let’s be honest, we’ve all been a little obsessed with AI lately. Whether you’re using ChatGPT to draft a tricky email or asking Claude to summarize a dense report, it’s clear this technology is more than just a passing fad. It’s powerful, it’s intuitive, and it’s already changing how we work.
But here’s the thing: for the insurance industry, what we’re seeing now is just the warm-up act. The main event is something called “agentic AI,” and it’s poised to quietly rewire how we do business by 2026.
Now, I know what you might be thinking. "Another AI buzzword? The insurance world moves at a glacial pace!" And you’re not wrong. We’re a cautious bunch, and for good reason. But according to the experts, this isn't just hype. It’s a slow, deliberate, and unstoppable shift.
So, What Exactly Is This "Agentic AI"?
If you think of today's chatbots as super-smart librarians who can find any piece of information you ask for, then think of agentic AI as a junior associate. You don't just ask it questions; you give it tasks. It can plan, make decisions, and actually do things on its own, all while you keep an eye on it.
It’s the difference between asking, “What are the steps to process a simple auto claim?” and saying, “Process this simple auto claim.” The AI would then access the data, follow the protocols, and complete the task. Big difference, right?
Inci Kaya, a research manager at IDC Financial Insights, really nailed it. She says carriers know what they want and aren't getting distracted by "all the shine and gloss." They’re taking a "steady and clear-eyed, deliberate" approach. It might feel slow, but it’s intentional.
Instead of flipping a switch and going "full-blown AI," most companies are running pilot programs. They’re testing it out on specific tasks, seeing what works, and learning as they go. It’s less of a revolution and more of a careful evolution—think "predictive analytics plus a sprinkle of AI."
From AI Sidekicks to AI Agents
Right now, a lot of the talk is about AI "co-pilots." These are tools, like a souped-up assistant, that help us do our jobs better and faster. Franklin Manchester, who’s the principal global insurance advisor at SAS, has a fascinating take on this. He believes that by next year, we could see these generative AI assistants start to make traditional policy admin systems look a little… well, old-fashioned.
Think about it. Policy admin systems cost a fortune to maintain. But what if you had a co-pilot that could connect directly to your data and interact with it, just like the admin system does? Manchester points out that sooner or later, a CEO is going to realize they don't need the expensive, clunky system to get the job done.
That’s the co-pilot stage. But Alex Sion, head of financial services at Blend360, thinks that by 2026, the real forward-thinking carriers will be moving on to the next level: full-blown agentic AI.
He says that while generative AI "turbocharged the productivity of humans," agentic AI will change the processes themselves. You’re not just making a human more efficient; you’re skipping human steps or replacing them with a new, autonomous workflow run by AI agents (with human oversight, of course).
For an industry as paper-heavy and process-driven as ours, this is a massive deal. Sion calls it a "massive unlock." It’s not hard to imagine a future where our roles are either dramatically supported by AI or, in some cases, even handled by it.
Okay, But Are We Actually Ready for This?
This all sounds great in theory, but let’s get real. Is the industry really prepared for AI agents to start running tasks?
Manchester is a bit skeptical about the timeline. He believes most insurers in the U.S. and globally won't be truly ready for agentic AI until late 2026. And even then, he predicts it will start with "low-downside-risk use cases."
In other words, don't expect an AI to be making the final call on a complex, multi-million dollar commercial claim anytime soon. That kind of human oversight isn't going anywhere.
Instead, Manchester suggests we flip a common phrase on its head. We always talk about having a "human in the loop." He wants us to think about having "AI in the loop." The idea is to use agentic AI for specific steps in a long process where the risk of something going wrong is low.
He paints a great picture: imagine agentic AI functioning like an entry-level claims processor. It facilitates the process, handles the routine stuff, but the most experienced person—the human—is still the supervisor with final authority. It's about augmenting the team, not replacing the manager.
First, We Need to Tidy Up the House
Before we can even dream of handing tasks over to AI, there’s some serious "housekeeping" to be done. Inci Kaya brought up a couple of critical points here.
First, there’s the data. AI is only as good as the data it’s trained on. If your data is a disorganized mess, your AI will be, too. Carriers need to get serious about structuring their data and setting up clear governance rules. As Kaya puts it, "Without those things, we’re just talking; we’re just playing pretend." This isn’t a one-year project, but it’s step one.
Second, there's the human side of the equation. If AI starts handling the entry-level tasks, how do new people learn the ropes? That hands-on experience is how generations of underwriters and claims professionals have developed their expertise.
Kaya worries that newcomers are "being deprived of the opportunity to learn hands-on." Sure, AI tools can help them, but she's not convinced a co-pilot can replace the intuition an underwriter builds over a 20-year career. That's a huge question mark we need to address with serious upskilling and new training models.
Your Customers Are Already Using AI—And That Changes Everything
Here’s what might be the biggest catalyst for change: our customers. While the industry is cautiously dipping its toes in the water, consumers are diving in headfirst.
People are already using tools like ChatGPT and Gemini for everything, including researching insurance. Alex Sion believes this shift in consumer behavior will force the industry to move faster. The days of just relying on Google are fading. Now, a customer can have a detailed, 500-question conversation with an AI about the nuances of a policy.
Franklin Manchester predicts that by 2026, half of all U.S. insurance consumers will use AI tools to research and shop for policies. That’s a staggering number. It forces us to completely rethink digital sales, marketing, and distribution.
But there’s a dark side to this, too. As customers get savvier with AI, so do the fraudsters. Manchester warns that we’re already seeing an alarming number of fraudulent claims using AI-generated forgeries—fake documents, videos, and emails.
The National Insurance Crime Bureau estimates that about 10% of claims have some element of fraud today. Manchester’s prediction? That number is going to double to 20% by 2026, all thanks to the rise of generative AI. It's an arms race, and we need to be prepared on the fraud detection front as well.
So, What's the Bottom Line?
Despite the challenges—the data cleanup, the workforce training, the threat of fraud—the overall feeling is one of optimism.
Kaya notes that the conversation at industry conferences has shifted. A year ago, everyone was worried about AI bias and "hallucinations." Now, the focus is on "how can we capitalize on this more solidly."
The readiness is there. The skepticism is turning into confidence. It’s now a matter of getting the internal strategy right: identifying the best use cases, getting the data house in order, and maybe even appointing a Chief AI Officer to bridge the gap between business and tech.
It won't happen overnight, and it won't be a dramatic, Hollywood-style takeover. It’s going to be a steady, piece-by-piece integration. But make no mistake, it’s happening. And by 2026, our new AI colleagues will be ready for their first real assignments.
