From tracking three major insurance technology conferences in the past 12 months, the Datos ILTF was the first where the dominant conversation shifted from “Does AI work?” to “How do we reorganize our operating model around it?” That distinction matters more than any individual statistic from the event. The industry has moved past proof-of-concept validation and into the harder problem of organizational design.
On April 22-23 at the InterContinental Boston, more than 100 carrier technology executives and 35 vendor sponsors gathered for the 2026 Insurance Leaders Technology Forum. Datos Insights used the event to introduce the “Intelligent Insurer Operating Model,” a framework built on survey data from 36 senior carrier technology leaders. The accompanying report, “Beyond the Horizon: The Dawn of the Intelligent Insurer,” published May 7, articulates the case that linear insurance workflows must give way to coordinated AI-human execution across the value chain.
Most coverage of the ILTF fixated on the headline statistic: 61% of carriers now have AI in production, up from 37% a year earlier. That number is real and significant. But the operating model framework deserves closer scrutiny, both for what it prescribes and for how well it maps to what Travelers, Chubb, and AIG have actually built.
The Five Pillars of the Intelligent Insurer
Tim Baum, Datos Insights’ lead analyst on the framework, structured the model around five pillars. Each addresses a different failure mode that has stalled carrier AI programs to date.
Pillar 1: Scale Without Headcount Growth. The framework positions agentic AI as the mechanism for expanding capacity without proportional staff increases. “Digital employees” absorb volume, repetition, and data processing while “biological employees” move up the value chain into judgment-intensive roles. Baum references Nvidia’s Jensen Huang, who has described running 42,000 biological employees alongside a projected workforce of “hundreds of thousands of digital employees.” The parallel to insurance is direct: underwriting shops that process 500 submissions per month with 20 underwriters could theoretically handle 2,000 with the same headcount if AI agents manage intake, data extraction, and preliminary risk scoring.
Pillar 2: Intentional AI, Not Experimentation. Operating model design must precede pilot programs. This is Datos Insights’ answer to the industry’s most persistent AI deployment failure: dozens of proofs of concept that never reach production. The ILTF survey data supports the diagnosis. Despite 61% reporting production deployment, 70% of carriers spend under $500,000 annually on AI projects. A single senior data scientist in New York commands $250,000-$350,000 in total compensation. These are experimentation budgets running inside production environments, not transformational investments.
Pillar 3: Reimagine Work, Don’t Digitize It. “Automating a broken process produces a faster broken process,” as Baum framed it at the conference. This pillar targets the bolting-on pattern we documented in our analysis of the ILTF’s claims-to-underwriting pivot: most carrier AI deployments layer technology onto existing workflows rather than redesigning the workflows themselves. Four of the top five deployed AI use cases at ILTF carriers fall within document processing: reading, extracting, summarizing, and classifying. These are important operational capabilities, but they digitize existing processes rather than reimagining them.
Pillar 4: Business-Led, Tech-Enabled Transformation. “The business composes the symphony and the CIO conducts the orchestra.” This is the organizational governance pillar, and it addresses a structural problem that recent survey data has quantified. A May 2026 survey of 344 senior executives and 809 employees found that 55% of insurers lack clarity on AI ownership within their organizations, and the industry allocates 72% of AI investment to technology and infrastructure while directing only 28% toward change management. Among the 10% of P&C insurers that have successfully scaled AI, the differentiator was not better technology; they were nearly four times more likely to invest in change management beyond basic training.
Pillar 5: Speed as Competitive Advantage. The framework’s most operationally specific claim: competitive separation will be determined by response time, not strategic positioning. Minute-level underwriting turnarounds versus day-long cycles. The ILTF showcased Jointly AI Broker, launched in February 2026, which completes commercial package quotes in 35 to 45 minutes from initial contact to bound policy as an illustrative case.
Production Deployment Doubled, But Scope Varies Widely
The 37%-to-61% production deployment jump in 12 months is the ILTF’s most-cited data point, and it deserves disaggregation. The Datos Insights survey polled 36 senior carrier technology leaders, a small but senior sample that skews toward larger, technology-forward organizations. Other survey instruments tell a more nuanced story.
The same month, Grant Thornton’s 2026 AI Impact Survey of 100 insurance executives found 62% rating their AI maturity as “scaling across multiple functions,” broadly consistent with the ILTF production number. But the governance picture complicates the optimism: 44% cited governance and compliance challenges as the primary cause of AI project failure or underperformance. Only 24% said they were very confident their organization could pass an independent AI governance review within 90 days.
Our earlier analysis of the 82% adoption vs. 7% scalable success gap documented a similar pattern from Sedgwick data in claims deployments. High adoption rates mask thin deployment. The question is not whether carriers are using AI in production but whether their production deployments are load-bearing or peripheral.
The ILTF survey data illuminates this distinction. Only 8% of respondents believe they currently lead their peers in AI capability. Yet 70% expect at least moderate competitive advantage within three years. That confidence-competence gap is precisely the problem the Intelligent Insurer framework aims to solve: carriers need an organizing principle for converting scattered production deployments into coordinated operating model transformation.
| Survey Source | Sample | AI in Production | Successfully Scaled | Governance-Ready |
|---|---|---|---|---|
| Datos Insights ILTF 2026 | 36 senior tech leaders | 61% | 8% self-assessed leaders | Not measured |
| Grant Thornton 2026 | 100 insurance executives | 62% scaling | 52% report revenue growth | 24% audit-ready |
| Insurance Business / Capgemini | 344 executives, 809 employees | 40% beyond POC | 10% fully scaled | 42% track zero metrics |
| Sedgwick (claims-specific) | Claims operations | 82% adoption | 7% at full scale | Not measured |
Reading across surveys, a consistent picture emerges: roughly 60% of carriers have AI running in production environments, but only 7-10% have achieved meaningful scale. The 50-point gap between deployment and scaled success is where the Intelligent Insurer framework positions itself.
Underwriting Displaces Claims as Primary AI Investment
The ILTF survey found underwriting AI live at 56% of surveyed carriers versus claims AI at 50%. That six-point gap represents a directional reversal. Through 2024, claims processing consistently led underwriting in AI deployment priority across industry surveys. ScienceSoft’s Q1 2026 insurance AI benchmarks corroborate the shift: claims processing, underwriting, pricing, and quoting accounted for 58% of disclosed generative AI use cases in Q4 2025, with underwriting-adjacent functions gaining share.
The economic logic is straightforward. Claims automation delivers cost savings: faster first notice of loss, automated damage estimation, reduced adjuster workload. Underwriting AI delivers revenue quality: better risk selection, tighter pricing accuracy, expanded capacity in complex segments. ScienceSoft benchmarks show AI driving a 67% decrease in manual underwriting tasks, reducing average decision time from days to minutes, with 90-99%+ accuracy rates in autonomous risk scoring for standard risks. For complex policies, the gains are smaller but still significant: 31% reduction in underwriting cycle time and 43% improvement in risk assessment accuracy.
The Intelligent Insurer framework positions this shift as evidence for Pillar 3: reimagine work rather than digitize it. Claims automation largely digitized existing processes. Underwriting AI, when deployed within the coordinated human-AI workflows the framework envisions, requires redesigning how risk decisions flow through the organization.
The Plug-and-Play Architecture Debate
One of the ILTF’s most consequential side conversations concerned vendor architecture. Kurt Diederich, CEO of Finys, articulated the argument in a Carrier Management executive viewpoint published the week after the conference. His framing: carriers should treat AI as a modular capability within a “plug-and-play operating model,” enabling evaluation, implementation, and replacement of components with minimal disruption.
The argument maps to the Intelligent Insurer framework’s second pillar, intentional design over experimentation. But Diederich adds a risk management dimension. The current AI vendor landscape exhibits “high volume of entrants, uneven differentiation, and ongoing capability leapfrogging.” He draws a historical parallel to the early 2000s internet proliferation, “where only a small percentage of vendors ultimately proved durable.”
The vendor concentration data supports this concern. Our analysis of IA Capital Group survey data found OpenAI sitting in 90% of carrier AI stacks with zero Google Gemini in production. When 70% of ILTF carriers spend under $500,000 annually on AI, they are overwhelmingly buying vendor solutions rather than building internal capabilities. A key vendor acquisition, product pivot, or failure would break AI-dependent workflows across much of the industry.
The plug-and-play model is partly a risk management strategy for exactly this scenario. But it creates its own tensions. Modular architectures require well-defined interfaces between components, standardized data formats, and governance frameworks that span multiple vendor relationships. These are organizational capabilities that the ILTF survey suggests most carriers have not yet developed.
Testing the Framework: What Travelers, Chubb, and AIG Have Built
The Intelligent Insurer Operating Model is prescriptive: it describes what carriers should do. The more useful question is how well it maps to what the most advanced carriers have actually implemented. Three distinct models have emerged from Q1 2026 earnings disclosures.
Travelers: Infrastructure-First at Scale
Travelers commits $1.5 billion annually to technology, with strategic AI spend more than doubling over eight years. The company has equipped nearly 10,000 engineers and data scientists with personalized Claude AI assistants through its Anthropic partnership, and more than 20,000 employees regularly use AI tools. The Travis digital quoting platform processes over one million transactions annually.
Against the five pillars, Travelers scores highest on Pillar 1 (scale without headcount growth) and Pillar 5 (speed). Claims call center population is down by a third, with locations consolidated from four to two. Over 50% of claims are eligible for straight-through processing, and two-thirds of eligible customers choose the STP option. CEO Alan Schnitzer describes “Innovation 2.0 at Travelers, powered by AI.”
Where Travelers deviates from the framework: it has invested heavily in Pillar 2 (intentional design) but through an infrastructure-led model rather than the business-led approach Pillar 4 prescribes. Travelers accumulated 65 billion clean data points over decades and built its AI strategy on top of that data asset. The transformation is technology-driven with business outcomes, not business-composed with technology execution.
Chubb: Disciplined Incrementalism With a Structural Target
Chubb’s approach under Evan Greenberg is the most disciplined of the three. The company targets 85% automation for major underwriting and claims processes over three to four years, with planned workforce reductions of up to 20% (approximately 8,600 of 43,000 employees) and projected run-rate expense savings of 1.5 combined ratio points upon completion. Roughly 85% of global gross written premiums already flow through fully digital or “significantly digitally enabled” channels.
Chubb maps well to Pillar 3 (reimagine work, don’t digitize it) because Greenberg has been explicit about restructuring workflows, not just adding AI to existing ones. The appointment of Kevin Rampe as the first Global Claims Officer in April 2026 is a structural prerequisite for the 85% claims automation target across 54 countries: you cannot coordinate AI-human workflows globally without a single point of authority.
Where Chubb departs from the framework: Pillar 5 (speed as competitive advantage) is secondary to Greenberg’s emphasis on executive knowledge and hands-on leadership evaluation. “You got to have firsthand knowledge. You can’t just be listening to others,” he told analysts. When your underwriting results already produce an 84% combined ratio, the risk calculus around AI deployment shifts. The downside of disruption is larger because there is more to protect.
AIG: Multi-Agent Orchestration
AIG represents the most architecturally aggressive implementation and the closest match to the Intelligent Insurer framework’s full vision. AIG Assist, the company’s internal AI platform, operates across eight lines of business with purpose-built agents for submission ingestion, risk evaluation, and pricing benchmarking. An orchestration layer coordinates agent activation, information sharing, and human oversight triggers.
The production metrics are specific: Lexington Insurance reported a 30% increase in quoted submissions, 55% reduction in time to quote, and approximately 40% increase in binding of submitted business. AIG processes four times more submissions with AI assistance. Agent autonomy has extended from less than one hour (Claude 2.0 baseline) to up to 30 hours of autonomous operation, a capability that CEO Peter Zaffino says evolved “at a faster pace than expected over the past nine months.”
AIG maps to all five pillars more completely than either Travelers or Chubb, but its approach also exposes the framework’s blind spot: governance. The 30-hour autonomous agent cycles create oversight challenges that the Intelligent Insurer model acknowledges in passing (organizational change management as a “critical parallel implementation requirement”) but does not structure into the five pillars themselves. Grant Thornton’s finding that only 24% of carriers could pass an independent AI governance review in 90 days suggests this omission is material.
| Framework Pillar | Travelers | Chubb | AIG |
|---|---|---|---|
| 1. Scale Without Headcount | Strong: 50%+ claims STP, call centers halved | Strong: 85% automation target, 20% headcount reduction | Strong: 4x submission processing with AI |
| 2. Intentional AI | Strong: $1.5B tech budget, 8-year ramp | Strong: structured 3-4 year plan | Moderate: rapid expansion across 8 LOBs |
| 3. Reimagine Work | Moderate: infrastructure-first, some workflow redesign | Strong: global claims restructure under single authority | Strong: multi-agent orchestration redesigns workflows |
| 4. Business-Led | Partial: technology-driven with business outcomes | Strong: CEO-led, firsthand evaluation | Moderate: technology architecture leads |
| 5. Speed | Strong: real-time quoting, STP | Secondary: discipline over speed | Strong: 55% time-to-quote reduction |
How the Framework Compares to McKinsey’s Agentic AI Roadmap
The Intelligent Insurer Operating Model is not the only framework competing for carrier attention. McKinsey’s April 2026 agentic AI analysis offers a different lens on the same transformation, projecting 10% to 90% productivity gains across insurance core system modernization phases.
The two frameworks operate at different altitudes. McKinsey focuses on the technology execution layer: agent libraries, orchestration meshes, the economics of agent reuse once initial development costs are absorbed. A human team of two to five people can supervise an “agent factory” of 50 to 100 specialized agents, per McKinsey’s organizational research. The biggest bottlenecks in core system modernization are not coding but the discovery, mapping, testing, reconciliation, and cutover loops that agents can compress.
Datos Insights operates at the organizational strategy layer: how the enterprise should be structured, who leads the transformation, and what competitive dynamics emerge from the transition. The five pillars are organizational design principles, not technology architecture patterns.
The frameworks converge on three points. Both emphasize modular, composable architecture over monolithic transformation. Both treat governance as a parallel requirement rather than an afterthought. And both identify speed as the primary competitive differentiator, though McKinsey quantifies it at the modernization-project level (testing productivity up 15-90%) while Datos Insights frames it at the enterprise-strategy level (minute-level response times versus day-long cycles).
Where they diverge is instructive. McKinsey’s framework is bottom-up: build the agent capabilities, demonstrate productivity gains, scale what works. The Intelligent Insurer model is top-down: design the operating model first, then deploy technology to fill it. Neither approach is wrong, but they produce different organizational behaviors. Bottom-up approaches generate faster initial wins but risk the fragmentation that Pillar 2 (intentional AI) warns against. Top-down approaches produce more coherent architectures but can delay production deployment while the organizational design work proceeds.
The consulting firm AI priorities analysis we published in March identified this same tension across Deloitte, Oliver Wyman, and McKinsey outlooks: all three firms agree on the destination but prescribe different sequencing for the journey. The Datos framework adds a fourth perspective to that set, one grounded in carrier survey data rather than consulting methodology.
The Governance Gap the Framework Underweights
The most significant limitation of the Intelligent Insurer Operating Model is what it treats as peripheral: governance. Organizational change management appears as a cross-cutting requirement, mentioned alongside the five pillars but not embedded within them. Given the survey evidence, governance may deserve its own pillar.
Grant Thornton’s 2026 survey quantified the problem with unusual precision. Among 100 insurance executives surveyed between February and March 2026, 52% reported AI-enabled revenue growth (15 points above the cross-industry average), and 62% reported improved decision-making from AI. These are real, measurable benefits. But 44% said governance and compliance challenges had caused AI project failure or underperformance. Only 24% expressed confidence they could pass an independent governance review within 90 days. The gap between benefit realization and governance readiness is wider in insurance than in any other sector Grant Thornton surveyed.
Mathew Tierney, Grant Thornton’s Global Insurance Practice Leader, captured the dynamic: “Insurance companies that are ahead have built AI governance into their operating model...with the specificity that regulators...are seeking.” That phrasing, governance “built into” the operating model, is precisely what the Intelligent Insurer framework does not do. It positions governance as a parallel workstream rather than a structural element.
The audit-layer failure analysis we published in April reinforced this finding: when carrier AI projects fail, the cause is typically not technology shortcomings but inadequate documentation, fragmented controls, and unclear accountability. These are operating model problems, exactly the domain the Datos Insights framework claims to address. Leaving governance as a side channel rather than a core pillar is the framework’s most consequential omission.
Why This Matters for Actuaries
The Intelligent Insurer Operating Model has specific implications for actuarial practice that extend beyond the technology adoption narrative.
Expense ratio assumptions require scenario modeling. If Chubb’s 85% automation target achieves the projected 1.5 combined ratio points of savings, that changes competitive positioning for every carrier in the same lines. Pricing actuaries should be building expense ratio scenarios around AI-driven cost reduction as a near-term planning assumption, not a long-range hypothetical. ScienceSoft benchmarks show large carriers achieving up to 30% improvement in portfolio performance and up to three-point improvements in loss ratios from AI-enhanced underwriting, numbers that directly affect rate level indications.
The operating model framework reshapes build-vs-buy analysis. The plug-and-play architecture that Diederich advocates and the Intelligent Insurer model implies has direct implications for actuarial vendor selection. Pricing models, reserve assumptions, and expense projections built around specific vendor capabilities may need to be modular, with documented assumptions about which AI components are load-bearing and what happens to financial projections if a component is swapped or deprecated.
Workforce modeling enters actuarial territory. When Chubb plans 8,600 fewer employees and Travelers reports call center staffing down by a third, the expense ratio impact flows directly into pricing and reserving work. But the transition period creates its own actuarial challenges: overlapping costs during implementation, productivity dips as workflows change (the J-curve pattern Morgan Stanley documented), and the risk that workforce reductions proceed faster than process redesign.
The framework demands actuarial involvement in operating model design. The financial feedback loops that actuaries understand, how underwriting decisions create reserve liabilities that affect capital allocation that constrains future underwriting capacity, are precisely the dynamics that need to be encoded into any AI-coordinated operating model. If carriers restructure workflows around coordinated human-AI processes following the Intelligent Insurer blueprint, actuaries need to be in the design room, not brought in after the architecture is set.
Governance readiness becomes an opinion-level concern. When 76% of carriers cannot demonstrate adequate AI governance on demand, and AI increasingly influences underwriting decisions that determine the risk distribution entering the book, reserve actuaries face new questions about whether AI-influenced portfolio composition changes are reflected in actuarial assumptions. The governance gap in actuarial practice becomes more acute as AI moves from peripheral tools to core operating model components.
The P&C combined ratio recovered to approximately 95% in 2025, the strongest underwriting performance in roughly a decade. Premium growth is projected at 4% for 2026. In a market where results are good but growth is decelerating, the Intelligent Insurer Operating Model describes the organizational transformation that separates carriers maintaining current performance from those building structural advantages for the next cycle. Whether Datos Insights’ specific five-pillar formulation is the right framework matters less than the underlying premise: the carrier AI conversation has moved from “should we?” past “does it work?” to “how must we reorganize?” That is an actuarial question as much as a technology one.