Your quote funnel was built for linear journeys.
Your next customer will zig-zag.

A customer might ask ChatGPT what kind of coverage makes sense. They might compare two options in conversation. They might open a structured quote experience, call a licensed agent for reassurance, and return later through a payment link. They might decide the product is right but the channel is wrong, at least for that moment.

The journey is not broken because it moved. It is broken when the context disappears every time it moves.

At Insurtech Insights NYC, Fletch convened a panel with Stripe, Profound, USAA, and Faye Travel Insurance around the question of how customers find and buy insurance when an AI assistant sits between them and the product. The throughline was simple: discovery, product presentation, payment, compliance, and handoff can no longer be treated as separate channel projects.

That is what makes this different from another chatbot initiative. A chatbot answers. An agentic journey has to understand, compare, route, authorize, and leave a usable record behind.

Walmart is building Sparky to synthesize reviews, make occasion-based recommendations, compare options, and help customers move toward a purchase. Target is pushing conversational shopping across multiple AI surfaces and allowing customers to build multi-item baskets. Accenture’s research describes the new market plainly: brands now have two audiences, the person and the AI agent acting on that person’s behalf.

Retail can move quickly because the object being sold is usually a product with a catalog description, a price, an inventory state, and a checkout path. Insurance has more moving parts. A quote is not a product card. Eligibility is not a recommendation badge. A bind is not just a completed cart.

Still, retail gives us the directional signal. The customer is going to expect the same continuity everywhere: ask once, compare clearly, know what happens next, and do not restart the journey because the surface changed.

In October 2025, Fletch completed what Insurance Innovation Reporter described as the first known insurance policy purchase executed entirely by an agentic AI system in a controlled production environment.

That matters because it was not a mock shopping flow. The transaction used AI to guide coverage choices, process payment, and complete policy binding. OpenAI models helped with the agentic experience. Stripe tokenized the payment credentials. The transaction happened with consumer consent and licensed producer oversight.

The important lesson is not that every insurance purchase should happen in chat. It is that chat, web, phone, payment, and bind can be parts of one transaction when the infrastructure carries the context across them.

The Fletch platform is designed around that idea: control which products surface, how they are presented, which disclosures appear, when to route to a licensed human, and how to measure the path from agent conversation to bound policy.

Start by accepting that your website is not necessarily the front door. Then ask a harder question: if an AI agent introduces your product, what does it need to know in order to represent it accurately and move the customer forward responsibly?

That means making availability, eligibility, coverage, price, exclusions, disclosures, and next steps legible to the systems that will carry the conversation. It means designing handoffs that preserve the customer’s answers and the carrier’s evidence. It means measuring the full journey instead of awarding all the credit to the first click or the last bind.

The old funnel is not dead. It is just no longer the whole funnel.

Your next customer will zig-zag. The carriers that win will make that feel intentional.