Who will buy the next insurance policy: A person or an AI agent?

The question for every carrier, agent and broker is not will this change distribution models but whether they are inside the new model or invisible to it.

To understand how to answer these questions, I have mapped the journey into five stages, from today through 2030.

Stage One: AI answers questions (now)

Instead of Googling or calling an agent first, consumers are using ChatGPT, Gemini, Claude, and Perplexity as primary research tools. AI referral traffic has grown dramatically with Adobe's April 2026 report showing internet traffic more engaged with AI, with March visitors 42% more likely to make a purchase. Why? Consumers have done their research and are ready to act. If your organization is not in these conversations, you're not even in the race.

Stage Two: AI reads products (next 6–12 months)

Next, making your policies readable and searchable by AI systems is critical to ensure your data is accessible in a format AI can understand. This is not about building a chatbot, instead it's about ensuring your policy data is structured so that when someone asks an AI assistant about coverage for a property, your products appear with accurate information on what you cover, where you operate and what makes you different. The good news is businesses of all sizes can now do this with ease.

Stage Three: AI initiates the journey (1-2 years into the future)

Here, AI moves from answers to action for consumers. Think of it as the difference between telling someone about a restaurant and making the reservation for them. A homebuyer asks their AI assistant to get coverage for a property, and the AI finds carriers, checks eligibility and begins the intake process. The consumer stays involved, but AI does the legwork. For carriers and agents, being searchable now means being reachable.

Stage Four: AI handles or refers (2026–2028)

For straightforward policies such as standard auto and renters, AI begins completing transactions using protocols like Agentic Commerce to allow purchases to be made programmatically. Amazon's Buy-for-Me beta feature is the example: a consumer asks for a product, and the AI agent buys it from another retailer on their behalf without the consumer leaving the Amazon app. Insurance is next.

For more complex risks, AI does something smarter than guessing—it recognizes its limits and brings in a human, either routing to the right specialist with full context or offering the consumer the option of expert guidance. This prevents a farm owner in an area with a history of hail damage from getting a generic quote from a national aggregator. AI identifies this quote demands local market knowledge and specialist appetite, and routes accordingly. The agent receives a warm, well-qualified opportunity at exactly the right moment, which is better for everyone.

Stage Five: AI negotiates and transacts (2028–2030)

All the major AI players are quietly developing always-on, personal AI agents designed to know who you are, what you need and what you own. The vision is a personal AI that handles the insurance conversation like a trusted advisor, knowing when to bring a human back in.

Stripe's Machine Payments Protocol provides the payment infrastructure: AI agents pay for services without creating accounts, entering card details or requiring a human at checkout. For many policy types, this will become the norm. For more complex risks, human involvement will remain central, either because the AI determines it is necessary or because the consumer chooses it.

To assess where your organization is, ask one simple question: if someone asked their AI assistant for coverage in your territory now, would you be in the answer? If you're guessing, you are likely at Stage 1, and working towards Stage 2 is the next task.

Expertise remains an advantage

AI agents are trying to answer complex, local, risk-specific questions on behalf of consumers. For that, they need depth no national carrier's database can easily replicate. They need the knowledge that only comes from decades of insuring a specific region through specialist carriers and independent agents. Local knowledge and specialization are not obstacles to agentic distribution. In fact, they are precisely what the new distribution layer seeks.

Discoverable, quotable and bindable

Discoverable means appearing in AI responses to consumer queries. Quotable means AI can retrieve accurate pricing and eligibility information for a specific risk independently. Bindable means a transaction can be completed or substantially progressed within the AI environment.

Most carriers are now discoverable. Getting to quotable and bindable requires a policy management system that exposes data in AI-friendly formats. The single most actionable step available to any carrier or agent today is to contact their policy management system vendor and ask: when will our policies be searchable by AI? If no clear answer is provided, that is important information. The more organizations that ask, the faster it becomes a priority.

The road ahead

Protocols are live. Consumer behavior is shifting. Google, Stripe and others are mapping how to make insurance purchases seamless and nearly automated. Those who act in the next 12 months will not just keep pace, they will establish the presence, integrations and trust signals AI agents require.

The expertise and insights of regional carriers and independent agents are not obstacles to this future. They are what the new distribution layer needs, and the window is open, for now.

Grahame Cohen is the Founder and CEO of Epoq, Inc., an InsurTech business dedicated to providing innovative products to the global insurance industry. For more than a decade, Grahame has assessed and implemented automation and AI-powered technologies that leverage trends in the insurtech market, working with major banks and insurance companies internationally. He holds a first-class honors degree in finance from City, University of London.

You can reach this contributor by sending an email to grahame@epoq.co.uk.

Opinions expressed here are the author's own.

(Featured image credit: Thaweerat/Adobe Stock)

Read additional thought leadership by this contributor:

• Common contract, risk transfer mistakes contractors make
• 3 Ways to mitigate risk when insurance isn't enough