A licensed carrier just began writing new business through a platform it doesn't own, doesn't fully control, and can't govern the way it governs its own website, and the industry still hasn't worked out what that means for the agent who used to own that customer relationship.
Liberty Mutual became the first major US carrier to sell auto and home insurance directly inside ChatGPT this year, after OpenAI opened its platform to customer-facing insurance apps on February 9.
The company said its in-chat quotes are generated by its own rating engine rather than the language model itself, which addresses one risk. It does nothing to answer a different one: what happens to the servicing relationship when the sale never touches an agent at all.
Plymouth Rock followed within weeks with its own ChatGPT quoting plugin, joining Tuio and Insurify, whose apps were among the first hundred approved on the platform. Whether a chatbot-originated quote gets routed to a servicing agent, assigned after the fact, or closed direct with no agent involved at all, is still unresolved across the industry.
With OpenAI counting more than 900 million weekly users as of February 2026, a pool no individual carrier's website comes close to matching, that ambiguity isn't a footnote. It's the whole story for anyone whose income depends on being the point of contact.
Friction with carriers is easing right as a new threat appears
Independent agents have spent the past year watching relations with carriers actually improve. First Connect's 2026 State of the Industry Report found year-on-year declines across every major measure of agent-carrier friction.
But the same report found that direct-to-consumer competition and rising digital expectations are opening a new front of pressure on the independent channel, with three-quarters of carriers still not using a third-party market intelligence provider to track any of it. Smoother carrier relationships mean little if the carrier's own new distribution channel quietly routes around the agent altogether.
Regulators aren't filling that gap either. The NAIC's Model Bulletin on the Use of Artificial Intelligence Systems by Insurers has now been adopted by 23 states and Washington, D.C., with a national AI evaluation tool piloting across 12 states, and New York's Department of Financial Services requires insurers to explain how AI factors into underwriting and pricing decisions under Circular Letter No. 7.
None of that oversight is built to answer the commission question. It's aimed at model governance and consumer protection, not at preserving anyone's place in the transaction.
That regulatory blind spot sits on top of an internal one. A GlobalData poll of insurance professionals published May 26 found close to one in four consider AI itself not yet ready for widespread use in the industry, pointing to unclear liability for AI errors and regulation that hasn't caught up.
If the industry's own professionals aren't confident in the guardrails, an agent watching a carrier plug straight into ChatGPT has reason to ask who's actually accountable when something goes wrong on a quote they never saw.
A new J.D. Power study helps explain why carriers see the risk as worth taking. The inaugural J.D. Power U.S. AI Insurance Experience Study found that 29% of auto and home insurance customers have used AI to research coverage, manage accounts, review coverage before a claim, or shop for a policy, and a third of those researching coverage are already choosing third-party AI tools over the ones their own insurer built.
Customers are showing up ready to transact with whatever's in front of them, agent or not.
The cost of standing still is measurable. A separate J.D. Power study on digital experience found that shoppers who used a chatbot or virtual assistant on an insurer's own site rated their satisfaction 132 points higher than those who didn't, and that price comparison tools nearly doubled the odds a shopper would consider buying, from 21% to 39%.
Only 11% of shoppers said they'd actually used a chatbot during the process, which is exactly the gap a platform with 900 million weekly users is positioned to close on a carrier's behalf, whether or not an agent is part of the transaction that results.
The same study found the behavior converts into real business: 37% of customers who used AI to research coverage changed their policy based on what it told them, and 42% of those who used AI to shop for a new policy completed a purchase.
Consumer appetite for insurer-side AI is rising too. Insurity's 2026 AI in Insurance Report found support for insurers using AI to improve service nearly doubled to 39%, up from 20% a year earlier, though comfort still has real limits: only 22% of consumers said they'd be okay with AI filing a claim for them, and just 16% would let it renew or cancel a policy unsupervised.
Tony Soloman, director of insurance intelligence at J.D. Power, said the shift echoes earlier technology waves.
"Just as we've seen with the internet and mobile apps, AI is rapidly becoming a critical conduit to key policy shopping, research and account management decisions among auto and home insurance customers," he said.
Soloman added that insurers now have to build tools that deliver a comprehensive, helpful user experience on their own platforms while tracking how their data gets interpreted by outside language models that customers are already choosing instead, a job that says nothing about where that leaves the agent originally meant to guide the customer through it.
Most customers still haven't gone near AI for any of this. Some 71% haven't used it for coverage research, 75% haven't used it for quotes, 87% haven't used it for claims, and 70% haven't used it for account servicing, with unfamiliarity, habit and distrust cited most often. That gap is exactly the window agents still have to make the case for a human relationship before carriers finish deciding whether they need one at all.
