42% of P&C Insurers Never Measured AI Outcomes: Inside the Capgemini Measurement Gap

From tracking quarterly AI disclosures across the top 20 carriers for the past 18 months, the gap between what carriers say on earnings calls and what they actually measure internally has become the clearest leading indicator of which deployments will scale. The Capgemini World Property and Casualty Insurance Report 2026, released May 5, puts global survey data behind that pattern. Based on interviews with 344 senior P&C executives, surveys of 809 insurance employees, and input from 1,113 policyholders across the Americas, Europe, and Asia Pacific, the report identifies a stark divide: a top 10% of "intelligence trailblazers" generated 21% higher revenue growth and 51% greater share price increases from 2021 to 2024. Meanwhile, 42% of insurers admit they have never measured AI outcomes at all.

Most coverage of the report repeats the 42% headline and moves on. This article dissects what the Capgemini methodology actually reveals, benchmarks its findings against carrier-specific Q1 2026 earnings disclosures, and proposes a measurement framework that actuaries can use to define the KPIs their organizations are missing.

What the Capgemini Methodology Covers

The World P&C Insurance Report is now in its 19th edition, which gives the Capgemini Research Institute a longitudinal baseline that most single-year surveys lack. The 2026 edition draws on three distinct data streams: executive interviews with 344 C-suite and senior P&C leaders, employee surveys capturing 809 respondents across underwriting (209), claims (200), customer service (200), and agent (200) roles, and a voice-of-the-customer survey of 1,113 policyholders fielded from December 2025 through February 2026.

The executive interview sample spans major markets in three regions, giving the findings a global footprint that single-market surveys (like the Insurity U.S. consumer data or Grant Thornton's U.S.-weighted AI impact survey) cannot replicate. The employee survey is particularly valuable because it captures the operational reality at the workflow level rather than the strategic narrative from the C-suite. When 47% of employees using AI tools report their workday has not changed after 18 months of deployment, that contradicts the productivity narrative that executives present in analyst calls.

The Capgemini Research Institute's lead for financial services, Luca Russignan, puts the structural problem plainly: "Insurers allocate 72% of AI investment to technology, while only 28% goes to change management, the factor that determines whether people use it effectively." That 72/28 split is the single most consequential finding in the report, because it explains why the 42% measurement gap exists. Carriers are buying tools without building the organizational capacity to evaluate whether those tools produce results.

The Intelligence Trailblazer Profile: What the Top 10% Do Differently

Capgemini segments its executive respondents into a top tier it calls "intelligence trailblazers," roughly 10% of the 344 executives interviewed. These carriers separate from the pack across three dimensions: strategy alignment, technology amplification, and organizational adoption. The performance gap is not marginal. Trailblazers achieved 21% higher revenue growth and 51% greater share price appreciation over the three-year measurement window from 2021 to 2024, compared with mainstream peers.

The behavioral differences that explain this gap are specific and measurable:

• Change management investment. Trailblazers are nearly 4x more likely to invest in change management programs that go beyond basic AI literacy training. Where most carriers default to "here is how to use the new tool" sessions, trailblazers redesign workflows around AI capabilities and restructure team incentives to reward adoption.
• Explainable AI infrastructure. Trailblazers are nearly 3x more likely to have deployed explainable AI systems that allow underwriters, adjusters, and actuaries to interrogate model outputs. This is not cosmetic transparency; it is the operational foundation for regulatory compliance under the NAIC Model Bulletin and the emerging state-level AI evaluation frameworks.
• Job description integration. Trailblazers are nearly 2x more likely to embed AI responsibilities directly into job descriptions rather than layering AI tools onto existing role definitions. This signals that AI use is a performance expectation, not an optional enhancement.

The common thread is that trailblazers treat AI as an organizational transformation rather than a technology procurement exercise. Kartik Ramakrishnan, CEO of Capgemini's financial services unit, frames it this way: "Trailblazers are proof that when carriers embed AI into their business strategy from the outset it elevates from an efficiency play into a true competitive advantage."

The 42% Measurement Vacuum: Anatomy of an Accountability Failure

The most striking finding in the Capgemini data is that 42% of P&C insurers have not measured AI outcomes in any systematic way. This figure deserves close examination because it exposes a structural problem that goes beyond technology readiness.

Consider what "not measured" means in practice. These carriers have invested in AI tooling, vendor contracts, cloud infrastructure, and data engineering pipelines. They have allocated headcount and budget. And yet they have no framework for determining whether those investments produced loss ratio improvements, expense ratio reductions, cycle time compression, or customer retention gains. The spending occurred; the accountability did not.

The 42% figure becomes more alarming when paired with two related data points from the same report:

• 55% report no clear ROI on AI initiatives. Even among the 58% who have attempted some measurement, more than half cannot identify a return. This suggests that the measurement frameworks in use are either poorly designed, measuring the wrong variables, or disconnected from financial outcomes.
• 55% say it is unclear who owns AI initiatives. Ownership ambiguity maps directly to measurement failure. When no individual or function is accountable for AI outcomes, no one is accountable for defining what "success" looks like, collecting the data to test it, or reporting the results. The measurement vacuum is, at its root, a governance vacuum.

From an actuarial standpoint, this is familiar territory. Actuaries have spent decades building measurement discipline around insurance operations: loss development triangles, expense ratio benchmarks, combined ratio targets, reserve adequacy testing. The AI measurement gap represents a failure to apply that same discipline to a category of spending that, for many carriers, now exceeds $50 million annually.

The Employee-Level Reality: 47% Report No Workday Change

The Capgemini employee survey introduces a perspective that executive-only surveys miss entirely. Across 809 insurance professionals in underwriting, claims, customer service, and agent roles, the findings reveal a workforce that is largely disengaged from the AI transformation their executives describe on investor calls.

Employee Survey Finding	Percentage	Implication
Workday unchanged after 18 months with AI tools	47%	Adoption without workflow redesign
"Very clear" on AI's role in their work	14%	Communication failure from leadership
Cite job security as top concern	43%	Trust deficit slows voluntary adoption
Time spent on cross-team collaboration	49%	Workflow fragmentation persists

The 47% figure is particularly telling. If nearly half of employees equipped with AI tools see no meaningful change in their workday after 18 months, the tools are either poorly integrated into existing workflows or solving problems that employees did not have. Either way, the productivity gains that carriers project in analyst presentations are not materializing at the operational level for a large share of the workforce.

Only 14% of employees say they are "very clear" on how AI fits into their role. This communication gap compounds the measurement problem. If employees do not understand what AI is supposed to help them accomplish, they cannot provide the feedback loop that measurement systems require. The frontline signal that would tell leadership whether AI is working, where it falls short, and how workflows need to change is largely absent.

The 43% who cite job security as their top concern point to a trust problem that technology deployment alone cannot solve. Carriers that launch AI tools with messaging focused on "efficiency" and "automation" without addressing the workforce implications create resistance that undermines adoption metrics from the start.

The 72/28 Investment Imbalance and Why It Persists

Capgemini's finding that 72% of insurer AI spending goes to technology and only 28% to change management is the structural root of the measurement gap. This ratio persists for reasons that actuaries will recognize from other contexts where capital allocation decisions are driven more by what is easy to justify than by what produces results.

Technology spending is visible and budgetable. A carrier can present a board-level business case for a $20 million AI platform investment with specific vendor deliverables, timeline milestones, and feature specifications. Change management, by contrast, involves training programs, workflow redesign, performance metric restructuring, and organizational communication, all of which are harder to specify, harder to budget, and harder to show tangible outputs for in a quarterly review.

Vendor incentives reinforce the imbalance. AI vendors sell technology, not organizational transformation. Their sales process, pricing models, and success metrics are built around platform deployment, not around whether the carrier's employees actually use the platform effectively. This creates a natural pull toward technology spending that the carrier's procurement function, which is structured to evaluate technology products, reinforces.

The skills gap compounds the problem. Two-thirds (67%) of executives cite a shortage of AI skills as a barrier to scaling. Yet the training priorities reveal a mismatch: 86% of carriers focus on baseline AI literacy, 52% on governance and risk management, 46% on advanced skill development, and only 27% on incentive and workflow redesign. The skills investment ladder is bottom-heavy. Carriers are teaching people what AI is without teaching them how to work differently because of AI.

Russignan's assessment captures the consequence: "The technology is maturing, but the organizational conditions to absorb it are not yet keeping pace." This is the insurance industry's version of a well-documented pattern across sectors. Technology adoption without organizational adaptation produces tools that sit on top of existing workflows rather than reshaping them.

Cross-Referencing Capgemini Against Q1 2026 Carrier Disclosures

The Capgemini data is a survey-based snapshot. To test whether its findings hold against actual carrier behavior, we can benchmark the report's key claims against public Q1 2026 earnings disclosures and AI-specific announcements from major P&C writers.

The industry backdrop is favorable. U.S. P&C insurers posted their biggest Q1 underwriting profit in 25 years, with a combined ratio of 89.1% before policyholder dividends and $22.1 billion in aggregate underwriting gains. This means carriers have the margin cushion to fund AI initiatives. The question is whether they are tracking the returns.

Travelers provides the clearest measurement example. After partnering with Anthropic in January 2026 to deploy Claude assistants to nearly 10,000 engineers and analysts, Travelers has reported "significantly elevated levels of engineering excellence and meaningful improvements in productivity." Their expense ratio dropped 3 points to 28.5 in 2025, though the company has been careful not to attribute the full improvement to AI specifically. By January 2026, the deployment had expanded to over 20,000 users, with claims call center consolidation already underway. Travelers represents a carrier that both invests in technology and measures the organizational impact, placing it in or near the trailblazer category Capgemini describes.

Progressive illustrates the data maturity advantage. Progressive's telematics flywheel, with 21 million connected policyholders feeding real-time driving data into pricing models, demonstrates what Capgemini means by "very high data maturity," a status only 12% of insurers achieve. Progressive's AI measurement is embedded in its pricing cycle: telematics-informed loss ratios are tracked monthly and compared against non-telematics segments, creating a continuous feedback loop that most carriers lack.

Allstate's ALLIE platform shows the trailblazer model at work. Allstate's proprietary agentic AI system processes 10 million emails annually and generates one-third of the company's software code. The key difference from the Capgemini "mainstream" profile: Allstate built AI measurement into the platform from inception, tracking processing throughput, code quality metrics, and operational efficiency in real time rather than retrospectively.

These carrier examples are consistent with the Capgemini trailblazer profile, but they also highlight the selection bias in the data. The carriers that publicly disclose AI metrics are overwhelmingly the ones with positive results to report. The 42% that have never measured likely includes many mid-tier and regional carriers that are neither disclosing AI metrics to investors nor tracking them internally.

Where the 60% Stuck in Pilot Mode Are Failing

Capgemini reports that 60% of P&C insurers remain in the exploration or proof-of-concept stage with AI. Combined with the 42% measurement gap, this means a substantial share of the industry is running pilots with no framework for deciding whether to scale, pivot, or kill those pilots.

The pattern is consistent with what we documented in our analysis of the Sedgwick data showing 82% adoption but only 7% scalable success. The Capgemini data adds a critical dimension that the Sedgwick data lacked: the reason most carriers are stuck is not that the technology does not work. It is that they have not built the measurement infrastructure to know whether it works.

Three specific failure modes emerge from cross-referencing the Capgemini executive and employee surveys:

Pilot-to-production governance gap. Carriers run AI pilots with dedicated project teams, ring-fenced data, and executive sponsorship. When the pilot succeeds, the transition to production requires integration with legacy systems, compliance with enterprise data governance, and handoff to operational teams that were not involved in the pilot. Each of those transitions introduces friction. Without measurement criteria defined before the pilot begins, there is no objective basis for declaring the pilot a success or justifying the production investment. Our earlier analysis of the Grant Thornton data found that 76% of insurance leaders could not demonstrate adequate AI governance on demand, and this governance deficit is where pilot-to-production transitions stall.

Vendor-defined success metrics. When carriers outsource AI deployment to vendors, they often inherit the vendor's definition of success: model accuracy, processing speed, or throughput volume. These technical metrics do not map to the financial outcomes that matter to actuaries and CFOs: loss ratio impact, expense ratio lift, claims cycle time reduction, or customer retention. The vendor says the model is 94% accurate; the actuary needs to know whether that accuracy translated into a 0.3-point improvement in the loss ratio on the affected book of business.

Siloed measurement ownership. The 55% of carriers who say AI initiative ownership is unclear are, by definition, carriers where measurement ownership is also unclear. IT tracks uptime and system performance. The business unit tracks operational throughput. Finance tracks total spend. But no single function owns the end-to-end measurement of AI impact from deployment through financial outcomes. Without that unified view, the fragments of measurement that do exist cannot be assembled into a coherent picture of ROI.

An Actuarial Measurement Framework for AI Outcomes

Actuaries are trained to build measurement systems. Loss development triangles, experience studies, and reserve adequacy analyses are all frameworks for converting operational data into financial insights over time. The AI measurement gap the Capgemini report exposes is, at its core, a problem actuaries are equipped to solve.

Based on the Capgemini findings, the carrier disclosures reviewed above, and the Deloitte observation that organizations embedding KPIs into AI deployments achieve ROI within six to nine months, we propose a four-layer measurement framework:

Layer 1: Operational metrics (measured weekly). These are the real-time indicators that tell you whether the AI tool is functioning as designed. Processing throughput, model latency, error rates, exception volumes, and human override frequency. These metrics should be owned by the technology function and reported through existing operational dashboards. They answer the question: "Is the AI tool running correctly?"

Layer 2: Workflow impact metrics (measured monthly). These connect AI tool performance to business process outcomes. Claims cycle time by complexity tier, underwriting quote-to-bind time, FNOL-to-first-contact duration, documentation turnaround, and employee productivity (measured as cases handled per FTE rather than as self-reported satisfaction). These metrics should be owned by the business unit and compared against pre-deployment baselines. They answer the question: "Is the AI tool changing how work gets done?"

Layer 3: Financial impact metrics (measured quarterly). This is where actuaries add the most value. Loss ratio impact on AI-affected segments, allocated loss adjustment expense (ALAE) trends, expense ratio by functional area, and premium per employee. These metrics require the actuarial team to isolate AI-attributable changes from other variables (rate adequacy, mix shifts, loss trends, weather patterns) using the same credibility-weighting and trending techniques applied to rate indications. They answer the question: "Is the AI investment producing financial returns?"

Layer 4: Strategic value metrics (measured annually). Market share in AI-affected segments, customer retention rates, new business hit ratios, and competitive positioning relative to peers. These longer-horizon metrics test whether the operational and financial improvements translate into durable competitive advantage, the 21% revenue growth and 51% share price premium that Capgemini attributes to its trailblazers. They answer the question: "Is the AI investment creating strategic value?"

Layer	Frequency	Owner	Key Metrics	Question Answered
Operational	Weekly	Technology/IT	Throughput, error rate, override frequency	Is the tool running correctly?
Workflow impact	Monthly	Business unit	Cycle time, quote-to-bind, cases per FTE	Is AI changing how work gets done?
Financial impact	Quarterly	Actuarial/Finance	Loss ratio, ALAE, expense ratio, premium/employee	Is the investment producing returns?
Strategic value	Annually	Executive/Strategy	Market share, retention, hit ratio, peer benchmarks	Is AI creating durable advantage?

The critical design principle is that each layer feeds into the next. Operational metrics without workflow impact metrics produce vanity dashboards. Workflow impact without financial impact produces activity reports. Financial impact without strategic context produces isolated ROI calculations that do not inform capital allocation. The trailblazers Capgemini identifies likely operate something close to this layered approach, even if they do not describe it in these terms.

The Morgan Stanley Expense Ratio Benchmark: What 200 Basis Points Requires

Morgan Stanley's projection that AI could deliver 200 basis points of expense ratio improvement for P&C carriers by 2030 (from 30.4 to 28.5) provides a concrete financial target that the Capgemini data can be tested against. If 42% of carriers are not measuring AI outcomes at all, and another 55% of those who try cannot identify clear ROI, the path to 200 basis points narrows considerably.

The Morgan Stanley projection implies industry-wide adoption and measurement. Under the Capgemini data, the realistic scenario is bifurcated: trailblazers will capture disproportionate expense savings while the unmeasured majority will see AI spending show up as an expense increase rather than a ratio improvement. The 72/28 technology-to-change-management spending split suggests that most carriers will accumulate AI infrastructure costs (which hit the expense ratio) without achieving the operational efficiencies (which would offset them).

For pricing actuaries, this creates a segmentation challenge. Industry-wide expense ratio trends may mask a growing divergence between carriers that measure and optimize AI spending and those that do not. Rate indications based on industry aggregate expense data could understate the competitive advantage accruing to measurement-disciplined carriers while overestimating the efficiency trajectory of the industry as a whole.

The Data Maturity Constraint: Only 12% at the Top Tier

Capgemini reports that only 12% of insurers achieve very high data maturity. This figure directly constrains how many carriers can realistically implement the measurement framework described above, because robust AI measurement requires reliable, consistent, accessible data across business functions.

The relationship between data maturity and measurement capability is mechanical. A carrier that cannot unify its claims data across lines of business cannot measure AI impact across those lines. A carrier with inconsistent data definitions between underwriting and claims systems cannot track whether AI-improved underwriting accuracy translates into better claims outcomes. A carrier with fragmented customer data cannot measure whether AI-driven retention initiatives are working.

The 12% data maturity figure also explains why the trailblazer cohort is roughly 10%. The overlap is not coincidental: carriers with high data maturity are the ones with the infrastructure to both deploy AI effectively and measure whether it works. The 42% measurement gap is, in significant part, a data maturity gap.

Why This Matters for Actuaries

The Capgemini measurement gap has specific implications for actuarial practice across pricing, reserving, and enterprise risk management.

Rate filing documentation under ASOP No. 23. Actuaries who include prospective AI-driven expense improvements in rate filings need data to support those assumptions. If a carrier's own measurement infrastructure cannot demonstrate that AI investments have produced expense savings, the actuary's basis for projecting future savings is weakened. The 42% of carriers with no measurement data and the 55% with unclear ROI are carriers where prospective AI-related rate adjustments rest on assumptions rather than evidence. ASOP No. 23 (Data Quality) requires actuaries to assess the quality of data used in their analyses, and a carrier that cannot measure its own AI outcomes is, by definition, providing incomplete data for this purpose.

Reserve adequacy in an AI-affected environment. Claims AI deployments that accelerate FNOL processing and reduce cycle times should, in principle, produce faster case reserve development and more accurate initial estimates. Reserving actuaries at carriers deploying claims AI need to monitor whether development patterns are actually changing. The Capgemini finding that 47% of employees see no workday change suggests that, for many carriers, claims AI has not yet altered the operational cadence enough to affect development factors. Actuaries should be cautious about shortening development tails or reducing IBNR factors based on AI capabilities that the carrier's own employees report have not changed their workflow.

Enterprise risk management and capital allocation. The AI J-curve pattern, where implementation costs precede efficiency gains, means carriers in the early stages of AI deployment will see expense ratio pressure before they see improvement. For carriers that are not measuring AI outcomes, the J-curve never resolves: costs accumulate without the measurement feedback that would either validate the investment trajectory or trigger a course correction. ERM actuaries should flag AI investments without measurement frameworks as an operational risk comparable to any other material expenditure lacking performance monitoring.

Model validation under ASOP No. 56. When AI tools from vendors feed into pricing, underwriting, or claims decisions, the actuary's model governance obligations under ASOP No. 56 extend to the AI models themselves. The governance gap in actuarial practice becomes acute when the carrier has not even established baseline metrics for AI performance. An actuary cannot validate a model whose outputs the carrier does not track.

The Path Forward: From 42% to Universal Measurement

Closing the measurement gap does not require new technology. It requires organizational discipline applied to existing data. The Capgemini data points to three immediate actions that carriers can take before their next budget cycle.

Assign explicit AI measurement ownership. The 55% ownership ambiguity must be resolved first. Whether the responsible party is the chief actuary, the chief analytics officer, or a dedicated AI program manager, someone must own the end-to-end measurement of AI outcomes from deployment through financial impact. This assignment should be documented, resourced, and included in the executive's performance objectives.

Rebalance the 72/28 spending ratio. Shifting even 10 percentage points from technology to change management, moving from 72/28 to 62/38, would fund the workflow redesign, training, and measurement infrastructure that the Capgemini data shows trailblazers invest in. The trailblazers' 4x higher investment in advanced change management is the single strongest predictor of AI success in the dataset.

Mandate pre-deployment measurement criteria. No AI project should advance from pilot to production without predefined success metrics at each of the four layers described above. This is not a radical proposal; it mirrors the approval gates that carriers already use for rate changes, reserve methodologies, and capital allocation decisions. Applying the same discipline to AI investments closes the loop between spending and accountability.

The Capgemini report's subtitle calls this "the moment of AI truth." For the 42% of carriers operating without measurement, the truth is that they cannot distinguish working AI investments from failing ones. For actuaries, the opportunity is to bring the measurement discipline that defines the profession to the category of spending that will increasingly determine which carriers thrive and which fall behind.