Insurers race to build AI foundations beyond model sophistication

AI-focused companies captured 95.2% of insurtech investment in Q1 2026, and in P&C insurance that money is flowing to carriers that can move data cleanly across policy administration, claims, billing, document systems, and decisioning layers. A June 23 Insurtech Insights piece centers the same constraint. The market is rewarding systems that can turn AI into operating infrastructure, not isolated experiments.

Capital is chasing AI, but the quarter is more concentrated than it looks

Gallagher Re’s Q1 2026 Global InsurTech Report puts total insurtech funding at $1.63 billion, a level that underscores how much capital remains available for software tied to insurance operations. In the same report, AI-focused businesses raised $1.55 billion across 68 transactions, and every one of the ten largest Q1 funding rounds involved AI companies. The quarter included only the sixth early-stage insurtech mega-round on record.

Gallagher Re highlights AI liability insurance and the evolving cyber insurance market. The AI conversation is no longer confined to internal productivity projects. Insurers are now thinking about AI as both a driver of underwriting change and a source of new exposures, especially where AI systems alter decision-making, liability allocation, or cyber risk profiles.

Why architecture now matters more than the model itself

The practical issue is not whether an insurer can buy or build a strong model. It is whether that model can operate inside the carrier’s existing stack without breaking governance, duplicating data, or creating another one-off workflow. In P&C, that means the AI layer has to connect to underwriting files, submission intake, claims history, billing events, document repositories, and decisioning rules, then present that information inside the systems where underwriters and adjusters already work.

When that does not happen, the result is a pilot that looks good in a demo and disappears in production. When it does happen, the carrier gets operational reuse: the same governed data and controls can support underwriting, pricing, claims handling, and customer engagement without rebuilding each use case from scratch. That is where time-to-value and total cost of ownership start to improve, because the organization is not paying to stitch together separate AI tools for every workflow.

The consequences are concrete:

• If underwriting cannot see claims context, risk selection stays fragmented.

• If claims tools cannot surface policy terms and document history at the point of decision, adjusters keep switching systems.

• If governance is bolted on after deployment, autonomous actions become hard to audit across teams and lines of business.

• If the AI layer is outside the core workflow, the carrier ends up with recommendations, not execution.

How the main platforms frame the buying decision

Guidewire is the clearest software platform in the material because it is explicit about the operating layer. Insurers can build and govern agentic and predictive AI on an open trusted core, use prebuilt AI agents for underwriting, claims, and customer service workflows, and embed AI directly into policy and claims processes so role-specific intelligence appears at the moment of decision. That is a materially different proposition from a standalone model or a generic copilot bolted onto the side of the stack.

If the AI lives inside the workflow, the insurer can tie it to permissions, audit trails, and process controls that already exist in the core platform. This only works if the carrier is willing to modernize the core and align data models.

What the advisory voices are really saying

McKinsey holds that only a few insurers have extracted outsize value from AI, and that joining that group requires a strategic, comprehensive approach that rewires the enterprise. The value shows up when the insurer changes the way work moves across underwriting, claims, service, and decision support.

BCG takes that point further: insurers need to redesign core processes such as underwriting and claims so autonomous AI agents can become the primary execution engine under human oversight. That is a higher bar than efficiency gains from a copilot. It implies process redesign, explicit control frameworks, and a willingness to let AI execute routine steps while humans supervise exceptions and material decisions.

The 2026 report is the final installment in a three-part AI-focused series that began in 2024, when the focus was AI across the insurance value chain from primary distribution through claims.