Key Takeaways
- Accenture’s July 2026 report found 81% of insurers achieved at least 5% Gross Written Premium growth from AI and data initiatives, with 7% reporting GWP gains above 20%.
- Only 23% of surveyed insurers have scaled AI enterprise-wide, with legacy system integration (cited by 50% of respondents) and data quality (cited by 45%) the primary blockers, gaps the report links directly to slower top-line growth for the remaining 77%. Seven percent of insurers deploying AI are already reporting Gross Written Premium gains above 20%, according to Accenture‘s July 2026 report “How Insurers Drive Revenue by Deploying AI with Intent.” The broader finding, drawn from 263 senior executives surveyed across the Americas, Europe and Asia-Pacific, is that 81% reported at least 5% GWP growth from AI and data initiatives. Only 23% have deployed AI across the full enterprise, which means most of that growth is coming from isolated use cases, not structural capability.
Revenue Numbers, With Caveats
The survey, conducted between October and November 2025, attributes GWP gains to improved pricing models, better personalisation and more effective cross-selling. Underwriting and claims account for the bulk of early wins, where data tends to be cleaner and returns more measurable. Accenture’s parallel 2026 Pulse of Change survey found that 86% of insurance employees say AI tools increased their productivity, and 83% of insurers reported cutting underwriting turnaround times by at least 5%. The causal chain between those operational gains and GWP improvements is plausible, but it relies on self-reported executive data rather than independent measurement.
Why 77% Haven’t Scaled
The report’s central finding is the gap between localised wins and enterprise-wide deployment. Fraud detection models may deliver real savings in claims without those outputs ever feeding back into underwriting decisions. Distribution remains siloed. Data architectures built over decades were not designed to support a unified intelligence layer. The result, according to Accenture, is a pattern of successful pilots that stop short of systematic change.
Legacy system integration is the most commonly cited barrier, flagged by 50% of respondents. Integrating modern AI tooling with decades-old core systems typically means extensive custom development, data migration and higher project costs. Data quality and accessibility follow at 45%: insurance data is frequently siloed, inconsistent across business units and difficult to standardise for model consumption. Without reliable input data, even well-designed models produce outputs that erode rather than build confidence. As enterprises in adjacent sectors have found when deploying multi-agent AI at scale the integration layer is almost always where deployment stalls.
The Skills Bottleneck
Technical infrastructure is only part of the constraint. The report finds that 70% of insurers are running AI skills programmes, but only 14% have scaled them enterprise-wide. AI literacy stays concentrated in specialist teams rather than reaching the underwriters, actuaries and claims professionals who interact with model outputs daily. That gap produces hesitancy, inconsistent use and decisions that discount AI recommendations without being able to articulate why.
Accenture’s prescription goes beyond training: redesigning job roles, aligning incentives and creating hybrid business-technology functions so AI becomes part of the operating model rather than a specialist overlay. Ravi Malhotra, Accenture’s Global Insurance Industry Lead, is cited in the report as saying reinvention requires cultures where professionals “trust AI, challenge it, and use it to do their best work.” The premium on human skills in AI-era roles makes that workforce redesign both harder and more urgent.
Five Actions Accenture Recommends
The report sets out five strategic priorities for insurers trying to close the deployment gap. The first is tying AI initiatives directly to P&L outcomes rather than treating them as standalone technology programmes. The second is a two-speed data strategy: modernising legacy infrastructure over a longer horizon while building AI-ready data capabilities in the near term. The third is scaling AI skills beyond specialist functions into business-facing roles. The fourth is implementing formal AI governance frameworks, which 56% of respondents say they have already adopted. The fifth is framing AI as a business capability with specific revenue targets rather than an efficiency programme.
On governance, the report frames the 56% adoption figure as evidence that formal frameworks are becoming a baseline requirement for insurers in regulated markets, not a discretionary investment.
Agents as the Scaling Mechanism
The report identifies agentic AI as the most likely mechanism for pushing enterprise-wide adoption forward. 68% of respondents expect AI agents to reshape core insurance workflows, and 57% of insurance employees already interact with agents regularly, 13 percentage points above the cross-sector average, according to the report. In practice, that means agents handling initial claims intake, routing routine underwriting requests and managing first-line customer queries, with human staff retaining oversight on complex or high-value cases. Major banks deploying AI agents have reported productivity gains that map closely to what insurers are projecting.
Accenture also identifies agentic workflows as a practical response to the legacy integration barrier. Agents interacting with existing systems through APIs can extract value from legacy infrastructure without requiring a full replacement programme, which shortens the payback period for AI investment while longer-term infrastructure work proceeds.
The Cost of Waiting
The report’s competitive framing for the 77% still running AI in pockets is direct: most are using AI to run existing operations more efficiently rather than to build new revenue lines. Insurers with integrated AI can price more accurately, issue policies faster and personalise products at a granularity that fragmented deployments cannot match. Efficiency gains are real, but they do not compound the way structural product and pricing advantages do. The longer that holds, the harder the gap becomes to close.
Originally published at https://autonainews.com/insurance-ai-paradox-81-gain-5-gwp-only-23-deploy/
Top comments (0)