The embedded insurance market is projected to generate between $176 billion and $210 billion in gross written premium by 2026. Most of that growth is driven by API integrations at the point of sale: travel insurance bundled into a flight booking, device cover added at checkout, landlord insurance triggered by a rental agreement. This is embedded insurance as it exists today. It is distribution through context.
The next phase is different. By late 2026, industry analysts project that AI agents will search for insurance, compare policies, and execute purchases autonomously on behalf of customers. This is not embedded insurance. It is agent-distributed insurance, and it changes the economics of the market in ways that most carriers have not yet priced in.
How Agent Distribution Works
For an AI agent to purchase insurance on behalf of a customer, it needs four things: real-time access to rating engines, underwriting rules, product data, and binding capability. The agent submits a risk profile, receives quotes from multiple carriers, evaluates them against the customer's preferences (price, coverage breadth, excess levels, claims reputation), and binds the policy. The customer may approve the final selection or may have pre-authorised the agent to act within defined parameters.
BCG's 2026 research on AI-empowered insurance customers identifies three emerging economies. The first is the Assistance Economy, where agents deliver complete customer experiences. The second is Adaptive Customer Experiences, where interfaces personalise in real-time. The third is the most consequential for carriers: Agentic Twins that represent customers with delegated authority across their financial relationships, including insurance.
The critical point: agents need secure, user-granted tokens before they can initiate transactions across multiple institutions. This makes delegated-consent architecture the infrastructure layer that determines which carriers participate in the agentic distribution channel and which are excluded from it.
What Changes for Carriers
Three structural shifts follow from agent-distributed insurance.
Price transparency becomes total. Today, a customer comparing insurance quotes visits three or four comparison sites, gets tired, and picks one. An AI agent compares every available product in the market in seconds. It does not get tired. It does not anchor on the first quote. It optimises ruthlessly. For carriers that compete on price, this intensifies competition to a degree that comparison sites never achieved. For carriers that compete on coverage quality or claims experience, the question is whether agents can evaluate those dimensions as effectively as they evaluate price.
In the cases I have been involved in, the answer is: not yet. Most agentic insurance tools optimise primarily on price and coverage scope. Claims reputation, speed of settlement, and quality of service in distress situations are harder to quantify and harder for an agent to assess. This creates a temporary advantage for carriers that invest in making their quality metrics machine-readable. An insurer that publishes its average claims settlement time, customer satisfaction scores, and complaint ratios in a structured, API-accessible format gives agents the data they need to recommend on quality, not just price.
Distribution costs restructure. Traditional distribution through brokers costs 15 to 25% of premium in commission. Comparison sites charge per click or per acquisition. Agent-distributed insurance will have its own cost structure, likely lower per transaction but potentially higher in aggregate if agents drive higher switching rates. Carriers need to model the unit economics: what does it cost to be discoverable and purchasable by AI agents, and how does that compare to existing distribution costs?
The carriers investing in open, real-time API access to their rating and binding engines are positioning for this channel. Those relying on manual quote processes, PDF-based policy documents, or closed distribution partnerships will be invisible to agents. Invisibility in an agent-distributed market is equivalent to not existing.
Customer relationships become intermediated. When an agent manages a customer's insurance portfolio, the carrier's relationship with the customer becomes indirect. The agent decides which carrier to recommend, when to switch, and whether to renew. Brand loyalty, which has historically protected retention rates in personal lines, is replaced by algorithmic optimisation. An agent does not care about brand. It cares about outcomes.
This is the deposit franchise problem, translated to insurance. Just as banks face deposit erosion when agents optimise cash allocation, insurers face retention erosion when agents optimise coverage. The response options are similar: compete on rate (expensive and margin-destructive), become the agent's preferred platform (requires API investment and competitive product design), or build proprietary agentic services (requires careful conduct management to avoid recommending your own products over better alternatives).
The Regulatory Dimension
The FCA's Mills Review is directly relevant here. When an AI agent purchases insurance on a customer's behalf, is it performing a regulated activity? If the agent compares policies and recommends one over another, is that regulated advice? If it binds a policy, is it acting as an intermediary?
The current regulatory perimeter was designed for human intermediaries. An AI agent that performs the same functions raises questions that the FCA has acknowledged but not yet resolved. The Mills Review is examining whether AI systems could provide services functionally equivalent to regulated activities while remaining outside the regulatory perimeter.
For carriers, the practical implication is dual. First, any carrier building its own customer-facing AI agent that recommends or purchases insurance must treat that agent as subject to the same conduct obligations as a human intermediary. Consumer Duty applies to the outcome, regardless of whether a human or an algorithm produced it. Second, carriers accepting business from third-party AI agents need to understand who bears conduct responsibility when the agent's recommendation leads to a poor outcome for the customer.
These questions do not have settled answers. But carriers that wait for regulatory certainty before engaging with agent distribution risk being locked out of the channel when the rules crystallise.
Preparing for the Shift
The practical steps for insurers in 2026 are:
First, audit your API capability. Can an AI agent access your rating engine, product data, and binding process in real-time? If not, you are not in the agent distribution channel. This is a technology investment, but the strategic value is distribution access.
Second, make your quality metrics machine-readable. Claims settlement times, customer satisfaction data, complaint ratios, renewal rates. Structured, API-accessible data that agents can use to evaluate your products on dimensions beyond price. The carriers that do this first will attract agent-driven business on quality, not just cost.
Third, model the economics. What does agent-distributed business look like in your portfolio? What is the expected switching rate, the cost of API infrastructure, the impact on retention? Stress-test against a scenario where 10 to 20% of personal lines business migrates to agent distribution within three years.
Fourth, engage with the regulatory question. Do not wait for the FCA or EIOPA to resolve the intermediary question. Build your agent-facing products with conduct obligations baked in, so that when the regulatory answer arrives, your systems are already compliant.
The embedded insurance market took a decade to reach $200 billion. Agent-distributed insurance will move faster, because the infrastructure (open banking APIs, delegated consent, agentic AI platforms) already exists. The question for carriers is not whether this happens. It is whether they are in the market when it does.
*To discuss how the 90-Day AI Acceleration programme can help your organisation prepare for agent-distributed insurance, contact the Value Institute.*