Your Next Customer Will Be an AI Agent

By Jen Linton, CEO, Fenris — November 19, 2026
For two decades, our industry has chased the same goal: a quote-to-bind experience so seamless the customer barely notices the work behind it. We’ve built comparison sites, prefill, instant quotes and click-to-call, with each step shaving off friction. Agentic shopping is the next step, and it is arriving faster than most carrier roadmaps assumed.
What has changed is how consumers use AI. They aren’t asking chatbots for information anymore, they’re asking for answers (“find me a better rate on my car insurance and tell me what I’d give up”) and, increasingly, for action. Model Context Protocol (MCP), the open standard that lets AI assistants connect to outside systems, is the plumbing that makes it possible. But speed without accuracy will not work for insurance, an industry where operating margins are slim and price is the most important consideration for shoppers.
The industry is picking sides
In less than a year, agentic quoting went from novelty to earnings-call talking point:
- Insurify launched the first insurance app in ChatGPT in February, sending broker stocks down on disruption fears. It now runs an AI Agent Gateway with its own MCP server, but blocked Meta’s Muse agent for stripping coverage details from quotes.
- EverQuote published Agentic Barbarians at the Carriers’ Gates and launched Luva, an AI-native shopping platform.
- Allstate will launch a protection-focused personal agentic assistant on its ALLIE platform in early 2027. CEO Tom Wilson: “We will be available to all personal agents.”
- State Farm is putting OpenAI-powered tools in its agents’ hands, and Progressive CEO Tricia Griffith says a small number of policyholders are already experimenting with agentic shopping.
- Sigo Seguros opened its auto MCP server to ChatGPT, Grok and Muse.
Consumers are ahead of most of the insurance industry. In Insurify’s survey of 3,002 drivers, 42% have already used an AI assistant to shop for car insurance, including 60% of Gen Z. And 39% would let AI move them to a new insurer if it found a cheaper rate.
Our view: meet consumers where they shop
At Fenris, we see agentic shopping as a natural evolution, not a threat. Carriers and agencies that want to stay relevant have to show up where consumers are, and in the way they want to shop and buy. Blocking agents may be a reasonable short-term guardrail, but being unquotable inside the assistant your customer already trusts is a long-term problem.
Along these lines, here are three trends the industry will need to pay attention to:
- Agents will expose underpriced segments. Shopping friction has always hidden pricing mistakes. When an AI can check every carrier in seconds, at every renewal, that friction disappears. Price a segment too low and agents will steer that business to you in volume. Price it too high and they’ll walk it out the door. In the U.S., new business already runs loss ratios 10% to more than 30% higher than renewal business, and that’s at human switching speed. Switching is accelerating, too: J.D. Power reports the 30-day auto switching rate is at a record high, up 35% since 2021. Expect the gap to show up in loss ratios long before it shows up in marketing reports.
- Accurate customer data becomes the whole game. AI is not perfect. In the U.S., J.D. Power found that 37% of insurance customers who used AI for research changed their policy based on what it told them. Yet attorneys at McGuireWoods warn that AI models routinely apply the wrong state’s law and cite cases that don’t exist, all with the same confident tone. There is also no standard yet for what customer information an AI agent passes to a quoting platform. Quotes built on what a consumer happens to remember, or what an assistant infers, will be wrong in ways that surface at bind or, worse, at claim. A quote has to rest on high-quality third-party data that can be relied on, not information manufactured by an agent on the spot. Garbage in has always meant garbage out; in agentic channels, it means garbage at scale.
- Humans stay in the loop at the moment that matters. AI will handle most of the interaction, but most people still want a human option before they commit. Only one-third of consumers say they trust AI-driven insurance decisions, according to Insurity. In a survey for the Independent Insurance Agents & Brokers of America, consumers named losing access to a trusted insurance professional as their top AI concern. The winning model isn’t AI or agent. It’s AI-facilitated, with a human available at the decision point.
The Fourth Door: lead decisioning matters more
People, not algorithms, are the ones getting insured. That’s why smart lead decisioning remains essential, but the playbook is changing. For decades there were three doors: agent, phone and web. Now there is a fourth: the consumer’s personal AI assistant, the carrier’s own agent, or an aggregator’s MCP server. I expect it to eclipse the other three in record time. Each one carries different intent signals, conversion behavior and lifetime value.
Carriers will need to answer new questions in real time. Is this agent-originated shopper likely to bind? Will they stay past the first renewal, or will the same agent that brought them here take them somewhere else just as fast? Which prospects belong with a licensed human, and which can go straight through? Scoring every prospect to guide bidding, routing, binding and retention is the insurance problem Fenris has always worked on. We’re there to support that with quality insurance data, early segmentation and indices that enable profitable growth, no matter how the customer (or the customer’s agent) arrives in the funnel.
What we’ll see in the next six months
How we get to the fourth door will be something to watch and learn from, with a few clear leaders stepping forward. Here is where I expect the industry to go between now and mid-2027:
- Allstate won’t be alone. At least one more top-10 personal lines carrier will announce its own consumer agent or open an MCP server before mid-2027.
- Comparison sites will move from quoting to binding in the chat. Plugins that return estimates will become full quote-to-bind flows inside AI assistants, with a handoff to a licensed agent as the selling point.
- Today’s blocks will become tomorrow’s deals. Marketplaces that block general-purpose agents will sign access agreements that spell out how coverage details and disclosures are shown.
- “AI agent” becomes a high growth lead source. Expect fewer leads, but better ones, since the AI does the comparing first. Those shoppers may carry lower lifetime value, because the agent that brought them will shop them again at renewal, so carriers will need to track, bid on and price them separately from web and phone.
- Regulators will start asking questions. With the NAIC already piloting a tool to evaluate how insurers use AI, expect the first formal guidance on what an AI agent must show a consumer before a policy is bound.
- Retention will have to be re-earned at every renewal. When an AI agent can re-shop a policy in seconds, inertia stops protecting the book. Expect shorter policy tenure and more competition at renewal than at first purchase.
- Products with different pricing will start being developed specifically for the Agentic channel. An agent-sourced sale comes with different acquisition costs and different data quality than a web or call-center sale, and carriers will begin to priced for that difference.
The bottom line
Agentic shopping won’t replace the fundamentals of insurance distribution – it will expose who has them. The fourth door is opening faster than any channel before it. The winners won’t be the carriers with the cleverest chatbot, they’ll be the ones who know their customers, price accurately, and decide in real time who to pursue and how.
The (AI) agent will see you now!