Insurity Calls Out P&C Core System AI Hype: A Vendor-by-Vendor Delivery Audit

Insurity published a press release on May 5, 2026 challenging the P&C core system market's AI messaging directly, arguing that competitors claiming effort reductions still route delivery through system integrators. President Jatin Atre framed it as AI adding a line item to vendor and SI invoices rather than removing one. The claim is testable, and so is Insurity's own.

Key Takeaways

• Two testable claims: complex commercial lines product setup compressed from years to weeks, and competitors' "up to a 50% reduction" still depending on integrators billing at markup.
• $50 million invested in AI and R&D across the Andromeda and Borealis releases, on a base of over 400 cloud deployments reaching 22 of the top 25 P&C carriers.
• Duck Creek's 50% configurator claim is "initially delivered by Duck Creek Professional Services", so the reduction applies inside an engagement rather than eliminating one.
• Morgan Stanley projects 2026 post-AI operating margins at 14.7% against a 15.2% baseline as roughly $3 billion of implementation cost lands before $9.3 billion of savings.
• 68% of carriers use third-party AI and 18% cite third-party model risk as a challenge, on AM Best's survey of about 150 rated carriers and MGAs.

What Insurity Actually Claimed

Two claims are specific enough to test. First, that policy setup for complex commercial lines can move from years to weeks on AI-native architecture. Second, that competitors advertising up to a 50% reduction in effort still depend on integrators running months-long or years-long implementations.

The backing is a $50 million investment in AI and R&D through the Andromeda release of November 2025 and Borealis of February 2026, over 400 cloud deployments, and production capability across underwriting, policy administration, claims, and analytics. Insurity serves 22 of the top 25 US P&C carriers and 7 of the top 10 MGAs.

The claim needs one qualification Insurity did not supply. Setting up a new product on an existing Insurity deployment is a different exercise from migrating a carrier off a legacy platform. Years to weeks is credible for the first: a carrier already running Insurity's policy administration that wants to launch a specialty line or add a state. A greenfield implementation still carries data migration, integration with distribution and claims, filing alignment, and training. For an installed base of 400 deployments the first case is most of the addressable work, which is why the framing is defensible and also narrower than it reads.

The Four Vendors Are Not Selling the Same Thing

Guidewire launched ProNavigator on April 16, embedded in InsuranceSuite and InsuranceNow with role-specific guidance and audit trails. It is an assistant rather than a configurator. It makes frontline work faster and does not touch the implementation timeline, because a carrier using it still needs the same project to reach the platform. Guidewire's position rests on 570 insurers across 43 countries and more than 1,700 completed implementations, with the largest SI partner ecosystem in P&C, and the PricingCenter integration is where the release touches rating directly.

Duck Creek's Agentic AI Platform followed on April 28 with five layers spanning neuro-symbolic reasoning, orchestration, governance, an open gateway, and core data integration. The Agentic Product Configurator arrived a day later claiming up to a 50% reduction in requirement and manuscript generation effort, converting underwriting manuals and rating guides into implementation-ready configurations. That is exactly the task Insurity targeted.

The qualifying sentence sits in Duck Creek's own material: the configurator is initially delivered by Duck Creek Professional Services. The 50% applies to effort inside an engagement, not to removing the engagement, and the platform architecture still routes through services as the delivery mechanism.

EXL is a fourth model entirely. Revenue reached $570.4 million, up 13.8%, with insurance at $194 million and data and AI-led revenue at 60% of the total, up 28%. A carrier buying from EXL buys outcomes rather than a platform, and the capability leaves when the contract does.

The billing model is what connects all four. A core system transformation at a large commercial carrier commonly exceeds $50 million once services, migration, integration, testing, and change management are counted, and COTS platforms typically take three to five years. An SI partner billing hourly or on multi-year fixed fee does not benefit from a faster implementation; it benefits from a more capable one it can charge more for. Guidewire's ecosystem is simultaneously its moat and the structural reason timelines resist compression.

The J-curve makes the timing worse before it makes it better. Morgan Stanley projects 2026 post-AI operating margins dipping to 14.7% against a 15.2% baseline as roughly $3 billion of implementation cost flows through ahead of $9.3 billion of projected savings, with carriers paying for the new platform and the legacy system at once during migration.

The Test No Vendor Has Passed in Public

Insurity framed its challenge around commercial and specialty lines rather than personal lines, and that choice is the substantive part. Personal lines run standardized forms and predictable coverage structures, so a configurator has well-defined rules to parse. A mid-market commercial package can involve manuscript endorsements, multi-location schedules, layered attachment points, filings across dozens of states, and rating logic with hundreds of interacting variables.

Four capabilities separate an AI-native claim from AI-assisted traditional implementation, and no vendor has publicly demonstrated all four in a live commercial environment:

• Rate algorithm portability, translating proprietary rating including manual factors, experience modification, and schedule credits without loss of fidelity.
• Filing integration, maintaining a live connection to SERFF or state systems so rate and form changes reach production without manual reconciliation.
• Governance output, producing documentation sufficient for an appointed actuary to certify that filed rates match implemented rates.
• Multi-state consistency, holding one coverage form at different rate levels, deductibles, and endorsement availability across 50 jurisdictions at once.

The third is where the exposure concentrates, because a rate implementation error introduced by a configurator propagates systematically across states and products rather than appearing in one filing. Rate implementation errors are already among the more common market conduct findings when a human makes them.

Lock-in moves the wrong way as this matures. Weaving AI into the core layer raises the switching cost rather than lowering it: a carrier whose rating rules and product configurations are generated and maintained by vendor-specific AI cannot easily extract that intelligence for a competitor.

The vendors diverge sharply here. Duck Creek supports MCP and A2A through its gateway; Guidewire's ProNavigator is proprietary; EXL's capability never enters the carrier's platform at all; Insurity has not publicly committed to comparable open protocols. Finys CEO Kurt Diederich argued in Carrier Management for treating AI as a modular capability replaceable with minimal disruption, drawing the parallel to early-2000s internet vendors of which few proved durable.

The buyer side is not yet equipped to enforce any of this. AM Best's survey of about 150 rated carriers and MGAs found 60% expecting AI to transform their business within one to three years but only about 20% at an advanced implementation stage, with legacy system integration cited as a barrier by 41% and 53% describing themselves as cautious pacesetters. Only 13% felt very confident measuring AI ROI.

The last number is the one that binds: 68% use third-party AI solutions and 18% name third-party model risk as a challenge. A carrier that cannot price the risk of a vendor's model is not positioned to audit the vendor's claim about it, which is the condition Insurity's challenge assumes carriers can escape, and which hyperexponential's analysis of shifting vendor relationships identifies as the gap the forward-deployed engineer model tries to close.