AI for insurance distribution: pricing in every channel

Insurance used to be sold through a small number of channels: brokers, direct sales and, for personal lines, price comparison websites. Today a single product may need to quote on an aggregator, inside a partner's checkout, through a broker portal, in a chat window and, increasingly, through an AI agent acting for the customer.

Each of these channels needs the same thing: a correct price, returned quickly, with a clear explanation of what is and is not covered.

More channels

• Price comparison websites need real-time prices at high volume and within tight response times during peak traffic.
• Embedded insurance places a quote inside another company's customer journey, such as buying a bike, booking travel or renting a flat. Customers expect that step to be quick and simple.
• Broker and partner platforms need access to rates and rules through an API, often with their own field names and data formats.
• Conversational channels let customers ask for a quote in everyday language and ask follow-up questions.
• AI agents can act for a customer by collecting details, requesting quotes and comparing options through APIs.

Each channel used to mean a separate integration and, often, a separate copy of the pricing logic. Separate copies lead to errors and to different prices for the same risk.

One rating engine for every channel

A distribution setup that scales uses a single source of pricing, available in every channel:

• a real-time pricing API for aggregators, partners and internal systems;
• embedded forms for websites and apps that do not want to build their own journey;
• an MCP server, so that AI clients and agents can find a product's inputs and price it correctly.

When rates change, every channel updates at the same time because there is only one model.

What AI adds to distribution

AI adds two things.

First, quoting can happen in conversation. A customer can describe their situation in everyday language, and an AI assistant can map it to the product's inputs, request a price and explain the result. The assistant does not need to understand pricing. It needs a pricing engine that lists the required inputs and explains the result.

Second, small channels become worth serving. Niche products, small partners and trial propositions were often not worth a dedicated integration. When an agent can read a product's input schema and quote it through a standard interface, adding a channel costs much less.

What the pricing engine needs to provide

For AI-led distribution to be safe, the pricing engine behind it needs:

1. A clear input schema with named fields, types, allowed values and plain-English descriptions that an agent can follow.
2. Consistent results, so the same inputs always give the same price in every channel.
3. Clear decisions on whether a quote is valid, declined or referred, with the reasons.
4. Explanations, in the form of a trace of how the price was calculated, for customers, brokers and complaints teams.
5. Speed to handle aggregator traffic and live conversations.

The AI manages the conversation and the engine calculates the price. Keeping these two jobs separate is what keeps AI-led distribution compliant.

Getting started

If your products are priced in spreadsheets or inside an older policy administration system, every new channel becomes a project. Moving pricing into a dedicated rating engine that serves APIs, forms and AI agents from the same model turns a new channel into a configuration task.

Ptally builds and runs this engine for insurers and MGAs. Every model is available through the API and MCP from launch.