From comparing vendor demo timelines against actual go-live dates across a half-dozen P&C core system migrations, the gap between AI marketing and AI delivery remains the industry's most expensive blind spot. When Insurity published its May 5, 2026 press release challenging the entire P&C core system vendor market's AI messaging, it broke an unwritten industry rule: vendors do not publicly call out other vendors' delivery failures. Insurity's President Jatin Atre put it bluntly: "AI was supposed to reduce cost for carriers, not add a new line item to their vendor and SI invoices."

The timing was not accidental. In the six weeks preceding Insurity's announcement, Guidewire launched ProNavigator (April 16), Duck Creek unveiled its Agentic AI Platform and Agentic Product Configurator (April 28-29), and EXL reported Q1 2026 results showing AI-led revenue crossing 60% of total company revenue. Each vendor positioned AI as the central value proposition for carrier modernization. Insurity's challenge, in effect, asked a question that carrier CFOs have been asking privately: if all this AI capability exists, why do implementations still take years and cost millions in professional services?

This analysis audits the four major vendors' AI claims against actual delivery evidence, examines the professional services cost structure that AI was supposed to compress, and identifies the specific evaluation criteria actuaries should apply when their carriers face core system procurement decisions.

Insurity's Challenge: What They Actually Claimed

Insurity's press release made two specific, testable claims that distinguish it from typical vendor marketing. First, that policy setup for complex commercial lines can be compressed "from years to weeks" using AI-native architecture. Second, that competitors claiming "up to a 50% reduction" in effort still depend on system integrators billing at high markup rates for implementations that run months or years.

The company backed these claims with its own track record: $50 million invested in AI and R&D through the Andromeda (November 2025) and Borealis (February 2026) software releases, over 400 cloud-based deployments, and production AI capabilities across underwriting, policy administration, claims, and analytics. Chief Revenue Officer Sylvester Mathis stated that "while our competitors are now promoting new agentic applications, Insurity can point to similar and deeper capabilities that are already live."

The scale of Insurity's installed base gives these claims additional weight. The company serves 22 of the top 25 P&C carriers and 7 of the top 10 MGAs in the United States, backed by GI Partners and TA Associates. That breadth of deployment means any AI capability Insurity rolls out reaches a significant share of the commercial and specialty lines market without requiring net-new implementations.

Insurity's specific focus on commercial lines is strategically significant. Personal lines AI, where most vendor marketing concentrates, involves relatively standardized products with high-volume, low-complexity transactions. Commercial and specialty lines, with their manuscript policies, multi-state filing requirements, complex rating algorithms, and layered coverage structures, represent a far harder test of whether AI can genuinely compress implementation timelines.

The Vendor Landscape: AI Claims Side by Side

To evaluate Insurity's challenge fairly, we need to examine what each major competitor actually announced, what evidence supports their claims, and where the gap between announcement and production delivery sits.

Guidewire: ProNavigator and the Palisades Release

Guidewire launched ProNavigator on April 16, 2026 as part of its Palisades product release. The AI assistant is embedded directly in InsuranceSuite and InsuranceNow, providing role-specific insights for underwriters, claims adjusters, billing specialists, and customer service representatives. VP of Product Management Amy Mollin described it as "AI that just works, helping insurers move smoothly from evaluating AI to successfully adopting it in everyday operations."

Guidewire's approach differs fundamentally from Insurity's in one respect: ProNavigator is primarily an assistant, not a configurator. It surfaces information, provides context-aware guidance, and enforces role-based access controls with audit trails. These are valuable capabilities for frontline productivity, but they do not directly address the implementation timeline and professional services cost problem Insurity targeted. A carrier using ProNavigator still needs the same implementation project to get onto Guidewire's platform in the first place.

Guidewire's competitive position rests on scale: 570+ insurers across 43 countries, 1,700+ successful implementation projects, and what the company describes as the largest R&D team and SI partner ecosystem in P&C insurance. The Palisades release also bundled PricingCenter rating capabilities into PolicyCenter, a move with direct actuarial relevance for carriers evaluating the build-vs-buy decision for rating engines. But scale cuts both ways; Guidewire's massive SI ecosystem has a financial incentive to maintain, not compress, implementation timelines.

Duck Creek: Five-Layer Architecture and the 50% Claim

Duck Creek made the most architecturally ambitious announcements in the period. The company launched its Agentic AI Platform on April 28, built on Google Cloud and powered by Gemini models, with a five-layer architecture combining neuro-symbolic reasoning, agentic orchestration, governance controls, an open AI gateway, and core system data integration. CEO Hardeep Gulati declared that "agentic AI will redefine how insurance operates, enabling carriers to move from manual processes to orchestrated end-to-end decisioning."

The following day, Duck Creek announced the Agentic Product Configurator, which claimed "up to a 50% reduction in requirement and manuscript generation effort" by converting carrier documentation (underwriting manuals, rating guides, forms) into structured requirements and implementation-ready configurations. This is precisely the product setup timeline that Insurity targeted in its challenge.

The critical detail: Duck Creek's Product Configurator is "initially delivered by Duck Creek Professional Services." The 50% reduction applies to the effort required within a professional services engagement, not to eliminating the engagement itself. A carrier still pays for Duck Creek's services team to run the AI-assisted configuration process. As we analyzed in our deep dive on Duck Creek's platform architecture, the five-layer design is genuinely sophisticated, but the business model still routes through professional services as the delivery mechanism.

Duck Creek also cited Boston Consulting Group's projection of up to $80 billion in annual U.S. P&C market impact from AI adoption. That number, while directionally plausible, represents a total addressable market estimate across the entire value chain, not a specific ROI metric for Duck Creek's product configurator. Carriers should treat it as context, not as a promise.

EXL: The Services-Led AI Model

EXL's Q1 2026 results revealed a company that has fundamentally reoriented toward AI-led delivery. Revenue reached $570.4 million, up 13.8% year-over-year, with the insurance segment generating $194 million. The headline number: data and AI-led revenues grew 28% year-over-year and now represent 60% of total company revenue. Management attributed this to "scaled deployments of AI inside core client workflows."

EXL occupies a different position in the vendor ecosystem from Guidewire, Duck Creek, or Insurity. It is primarily a services company that builds AI into delivery rather than a platform company that sells AI-enabled software. For carriers, this distinction matters because EXL's incentive structure is aligned with demonstrating AI productivity gains (which justify premium pricing for services) rather than compressing implementation timelines (which reduce billable hours).

The 60% AI revenue mix is impressive as a corporate metric but ambiguous as a carrier benefit. A carrier paying EXL for AI-led analytics, claims processing, or underwriting support is buying outcomes, not a platform it controls. When the contract ends, the AI capabilities leave with EXL. This creates a different form of vendor dependency from core system lock-in, but dependency nonetheless.

Insurity's Own Position: Testing the "Years to Weeks" Claim

Insurity's most aggressive claim, compressing complex commercial lines product setup from years to weeks, requires scrutiny equal to what we have applied to competitors. The company's AI capabilities span real-time risk intelligence, advanced catastrophe modeling, intelligent submission scoring, document intelligence for submission classification, AI-enabled premium audit self-service, and agentic first notice of loss experiences. All of these are described as live in production, not future roadmap items.

The distinction between "product setup" and "full platform implementation" matters here. Setting up a new insurance product on an existing Insurity deployment, where the carrier already runs Insurity's policy administration and rating systems, is fundamentally different from migrating a carrier from a legacy platform to Insurity. The "years to weeks" claim likely applies to the former scenario. A greenfield implementation of Insurity's platform for a large commercial carrier would still require significant time for data migration, integration with existing distribution and claims systems, regulatory filing alignment, and user training.

That said, the distinction is meaningful. For Insurity's installed base of 400+ deployments, the ability to configure new products in weeks rather than months represents genuine value. A carrier already on Insurity's platform that wants to launch a new specialty line, add a state, or restructure a commercial product can theoretically move at the speed Insurity describes. For the broader market not yet on Insurity's platform, the timeline compression claim needs qualification.

The Professional Services Problem

The subtext of Insurity's challenge, and the structural issue that makes it resonate with carrier executives, is the professional services billing model that undergirds the core system vendor ecosystem. McKinsey's research on core system modernization documents that building a proprietary core system requires five to ten years of development and integration work, while COTS (commercial off-the-shelf) platforms typically deliver within three to five years. Even "small-scale" modernization projects run six months to a year.

For a large commercial lines carrier, a core system transformation commonly exceeds $50 million in total cost when professional services, data migration, integration, testing, and organizational change management are included. The system integrator ecosystem, which includes Accenture, Deloitte, Cognizant, TCS, and specialized insurance consultancies, derives significant revenue from these multi-year engagements. When a vendor claims "up to 50% reduction in effort," SI partners face a direct threat to their revenue base.

This creates a structural conflict of interest that Insurity's challenge implicitly exposed. Guidewire's "largest SI partner ecosystem in P&C insurance" is simultaneously a competitive moat and a drag on timeline compression. Partners who bill hourly or on fixed-fee multi-year projects do not benefit from faster implementations. They benefit from more capable implementations, which they can charge more for per hour while maintaining or extending the engagement timeline.

The J-curve in insurance AI implementation costs compounds this problem. Morgan Stanley projects that 2026 post-AI operating margins will actually dip to 14.7% versus a 15.2% baseline as $3 billion in implementation costs flow through before the $9.3 billion in projected savings materialize. Carriers investing in AI-native core systems are, in effect, paying twice during the transition: once for the new platform and once for the continued operation of legacy systems during migration.

AM Best's November 2025 survey of approximately 150 rated carriers and MGAs quantifies the adoption challenge. Sixty percent expect AI to significantly transform their business model within one to three years, but only about 20% report AI implementation at an advanced stage. The top barriers: data readiness (45%), security and privacy (43%), and legacy system integration (41%). That last number, 41% citing legacy system integration, is precisely the pain point core system vendors claim AI will solve. Yet 53% of carriers describe themselves as "cautious pacesetters rather than first movers."

Commercial Lines as the Litmus Test

Insurity's decision to frame its challenge around commercial and specialty lines, rather than personal lines, was deliberate and revealing. Personal lines core system implementations are difficult but fundamentally tractable: standardized ISO forms, high-volume rating algorithms, predictable coverage structures, and established distribution patterns. AI assistants and product configurators can meaningfully accelerate personal lines work because the domain rules are well-defined and relatively static.

Commercial and specialty lines present a categorically different challenge. A single commercial property policy may involve manuscript endorsements, multi-location schedules, layered coverage with varying attachment points, jurisdiction-specific filing requirements across dozens of states, complex rating algorithms with manual adjustment factors, and underwriting guidelines that vary by class code, industry segment, and individual account characteristics. The rating logic alone for a mid-market commercial package can involve hundreds of interacting variables.

When Insurity claims it can compress product setup for complex commercial lines from years to weeks, the implied capability is that its AI can parse underwriting manuals, rating guides, and regulatory filings to generate implementation-ready configurations without extensive manual translation by SI consultants. Duck Creek's 50% effort reduction on the same task represents a less aggressive but perhaps more defensible claim.

For actuaries, the commercial lines focus introduces a specific set of evaluation criteria that personal lines AI demonstrations do not test:

  • Rate algorithm portability: Can the AI correctly translate a carrier's proprietary rating algorithm, including all manual rating factors, experience modification calculations, and schedule credits, into the new platform's rating engine without loss of fidelity?
  • Filing integration: Does the platform maintain a live connection to SERFF or state-specific electronic filing systems so that rate and form changes flow through to production without manual reconciliation?
  • Model governance controls: When AI generates a product configuration, does the output include sufficient documentation for the appointed actuary to certify that filed rates match implemented rates? This is not a theoretical concern; rate implementation errors are among the most common findings in state market conduct examinations.
  • Multi-state consistency: Can the AI handle the same coverage form with different rate levels, deductible options, and endorsement availability across 50 jurisdictions simultaneously?

No vendor has publicly demonstrated all four capabilities in a live commercial lines environment. The vendor that does so first will have a legitimate claim to AI-native architecture rather than AI-assisted traditional implementation.

The Vendor Lock-In Calculus

Insurity's challenge carries an irony that the company did not address: replacing one vendor's lock-in with another vendor's AI-embedded lock-in does not solve the structural problem. When AI is woven into the core platform layer, as all four vendors are now doing, the switching cost increases rather than decreases. A carrier whose rating algorithms, underwriting rules, and product configurations are generated and maintained by vendor-specific AI cannot easily extract that intelligence for migration to a competitor.

Kurt Diederich, CEO of Finys, articulated this risk in Carrier Management, advocating for a "plug-and-play operating model" where carriers treat "AI as a modular capability" that can be "evaluated, implemented, and replaced with minimal disruption as conditions evolve." He drew a parallel to early-2000s Internet companies, noting that "only a small percentage of vendors ultimately proved durable." Carriers that treat AI vendor selection as a static investment, Diederich argued, risk accumulating technical and operational constraints that limit their ability to adapt.

The vendor approaches to openness vary significantly:

  • Duck Creek explicitly supports the Model Context Protocol (MCP) and Agent-to-Agent (A2A) protocol through its AI Gateway, positioning itself as a hub for multi-vendor agent orchestration. The architectural intent favors portability, though production validation is pending.
  • Guidewire offers a Jutro design system and developer tools that allow integration of external AI providers, but ProNavigator itself is a proprietary, Guidewire-only experience.
  • EXL operates as a services layer, which means the AI stays with EXL rather than being embedded in the carrier's platform. This is simultaneously the lowest lock-in risk (the carrier's core system is unchanged) and the highest dependency risk (capability evaporates when the contract ends).
  • Insurity has not publicly committed to open agent protocols comparable to Duck Creek's MCP/A2A support. Its AI is deeply integrated into its own platform, which maximizes performance for existing customers but raises questions about interoperability for carriers running multi-vendor technology stacks.

hyperexponential's analysis of AI-driven vendor-carrier relationship shifts identifies an emerging safeguard: the "forward deployed engineer" model, where vendors embed technical resources during the sales process to demonstrate working value before contract signing. This compresses the traditional RFP-to-pilot timeline from months to days and gives carriers empirical evidence rather than slide deck projections. In a market where rating model development can be compressed from three months to 30 minutes with the right tooling, carriers can demand proof of AI capability rather than accepting roadmap promises.

What Carrier CFOs and Actuaries Should Demand

Insurity's challenge, stripped of its competitive positioning, raises legitimate procurement questions that carrier decision-makers should formalize. Based on patterns we have tracked across core system transitions and BCG's three-phase AI transformation framework, here is what actuaries and technology leaders should require during vendor evaluation:

1. Separate AI Marketing from Delivery Evidence

For every AI capability a vendor demonstrates, ask: Is this in production with a named carrier, or is it a lab demo? The AM Best survey found that only 13% of carriers felt "very confident" in measuring AI ROI accurately. Vendors should provide reference customers willing to discuss actual go-live timelines, not just pilot outcomes. The gap between pilot success and production deployment remains the industry's largest credibility test.

2. Require Professional Services Cost Transparency

If AI reduces implementation effort by 50%, the professional services bill should decline by a comparable amount. Carriers should demand fixed-price contracts that reflect AI-assisted delivery, not traditional time-and-materials arrangements where the vendor captures all the productivity gain. Ask specifically: What percentage of total contract value goes to professional services versus software licensing? How has that ratio changed since AI capabilities were introduced?

3. Test Commercial Lines Complexity

Do not accept personal lines demonstrations as proof of commercial lines capability. Require the vendor to configure a multi-state commercial property or general liability product using AI-assisted tools during the evaluation. This single test will reveal more about actual AI maturity than any number of webinars or white papers. If the vendor cannot demonstrate commercial lines AI configuration in a controlled evaluation environment, the "AI-native" claim is aspirational rather than operational.

4. Evaluate Rate Algorithm Portability

For actuaries specifically, the most critical question is whether the AI can correctly implement your carrier's proprietary rating algorithms in the new platform. Request a rate reproduction test: provide the vendor with your current rating manual and a sample of policies, then compare the AI-generated rates to your production system's output. Any deviation requires explanation and resolution before contract signing.

5. Map the Lock-In Exposure

Ask each vendor: If we decide to leave your platform in five years, what data, configurations, and AI-generated artifacts can we export in a vendor-neutral format? Vendors committed to open protocols (MCP, A2A) should be able to articulate specific portability mechanisms. Vendors without open protocol support should explain their proprietary export capabilities and any contractual provisions for data portability.

6. Demand Governance Documentation

When AI generates product configurations, underwriting rules, or rating structures, the appointed actuary must be able to certify compliance with filed rates. Require vendors to demonstrate how AI-generated outputs are documented for regulatory examination, how changes are version-controlled, and how the audit trail connects AI recommendations to human approval. This is not optional; it is a regulatory necessity under ASOP No. 56 and increasingly under the NAIC's AI evaluation framework.

The Adoption Reality Check

The gap between vendor ambition and carrier readiness remains substantial. AM Best's survey data paints a sobering picture: while 60% of carriers expect AI to significantly transform their business within three years, only 20% have reached advanced implementation stages. The cautious-pacesetter mentality (53% of respondents) reflects the regulated reality of insurance: carriers cannot afford to be wrong about core system decisions that touch every policy, every claim, and every regulatory filing.

Workforce productivity gains provide the strongest evidence that AI is delivering measurable value. Sixty-three percent of surveyed carriers reported at least a small improvement in productivity and workforce satisfaction, with 11% reporting significant improvement. Efficiency gains in expense ratios and automation showed similar patterns (49% small improvement, 10% significant). These are real but incremental results, not the transformative compression that vendor marketing implies.

Perhaps the most telling statistic: only 18% of carriers cited third-party model risk as a challenge, despite 68% using third-party AI solutions. That disconnect suggests many carriers have not yet grappled with the governance implications of embedding vendor AI into core decision-making processes. When (not if) a regulator asks a carrier to explain how its AI-assisted rate implementation works, the answer "our vendor handled that" will not satisfy a market conduct examiner.

Why This Matters for Actuaries

Core system transitions have traditionally been technology decisions owned by CIOs and CTOs. The embedding of AI into rate implementation, product configuration, and underwriting rules changes that dynamic fundamentally. When the core system's AI generates rating structures, the actuary's certification responsibility extends to validating AI outputs, not just the filed rates themselves.

Three specific implications stand out:

Rate implementation risk shifts upstream. Historically, actuaries filed rates and product teams implemented them in the core system, with testing to verify fidelity. When AI compresses the filing-to-implementation pipeline, the actuarial team needs visibility into the AI's translation logic. A filed rate that is correctly approved but incorrectly implemented by AI creates the same regulatory exposure as a manual implementation error, potentially more, because the error may propagate systematically across states and products.

Model validation scope expands. ASOP No. 56 already requires actuaries to understand the models they rely on. When the core system itself becomes an AI model (generating configurations, recommending rating structures, flagging underwriting exceptions), the boundary between "the actuary's model" and "the vendor's platform" blurs. Actuaries evaluating core system vendors should demand access to the AI's decision logic at a level sufficient for independent validation, even if the vendor considers it proprietary.

Vendor selection becomes an actuarial opinion dependency. The appointed actuary's opinion on reserves, rates, and risk classification increasingly depends on the accuracy and governance of the core system's AI. A poorly governed AI-native core system creates downstream risks that flow directly into the actuarial opinion. Actuaries should have a formal role in core system vendor evaluation, focused specifically on rate fidelity, model governance, and regulatory compliance capabilities.

Insurity's public challenge to the vendor market, whatever its competitive motivations, surfaced a question that the industry needed to confront: if AI is truly transforming core systems, where is the evidence in shorter timelines, lower costs, and better outcomes? The vendors that can answer that question with production data rather than projections will define the next generation of P&C technology infrastructure. The rest will be remembered as marketing.