What Deloitte’s 2026 AI Survey Means for Insurance
By James W. Moore
Key Takeaways
Deloitte’s 2026 State of AI in the Enterprise survey found financial services carries the highest share of surface-level AI use of any industry surveyed, even as worker access to AI tools has roughly doubled. Insurance’s constraint isn’t buying AI. It’s building the process, governance, and workforce structures that turn access into operating capability.
Three findings from the report translate into three distinct leadership problems for insurance. Sovereign AI is becoming a regulatory and infrastructure question. Agentic AI is scaling faster than the governance built to control it. And automation expectations are outrunning any real redesign of the roles that used to train the next generation of underwriters and claims professionals.
Deloitte’s AI Institute surveyed 3,235 director to C-suite leaders across 24 countries and six industries between August and September 2025 for its 2026 State of AI in the Enterprise report. It is not an insurance study. The closest thing to one is the companion Financial Services cut, based on 573 leaders across banking, insurance, payments, investment management, and wealth management. Neither report is insurance specific. That is a feature, not a limitation.
One number from the Financial Services cut stands out. Forty-one percent of financial services organizations describe their AI use as surface level, meaning little or no change to existing processes, the highest share of any industry Deloitte surveyed.
That statistic is close to a direct answer to a question insurance leaders keep asking themselves. The industry has not been slow to try AI. Worker access to sanctioned AI tools in financial services doubled in a single year, from 30 percent to 62 percent. What has lagged is everything downstream of access: whether that access changes how underwriting, claims, and distribution actually get done.
Insurance rarely sets the pace on enterprise technology. It absorbs innovations after they have matured elsewhere, filtered through regulation, capital requirements, and legacy systems.
That is exactly why a cross-industry report is worth reading closely rather than skipping for something insurance-specific. It shows the pressures before they arrive, not after. And the pressures Deloitte documents are not abstract industry trends.
They map onto three distinct problems insurance leadership will have to solve separately: where AI infrastructure runs, how autonomously AI is allowed to act, and who is left to do the work AI has not yet touched.
Where AI Runs Is Becoming a Strategic Decision
As AI moves from isolated pilots into live workflows, organizations are discovering they have also inherited infrastructure decisions they never planned to make deliberately. Deloitte calls this sovereign AI, and in financial services it has moved well past a policy conversation. Eighty-six percent of financial services companies say sovereign AI is at least moderately important to their strategic planning, and 57 percent call it very important, well above the 43 percent cross-industry average on that same question. Eighty percent now factor a technology’s country of origin into vendor selection, and just over half say they build their AI stacks primarily with local vendors.
A regional carrier might assume this is someone else’s problem, a concern for multinational insurers managing books across US, EU, UK, and Asia Pacific regulatory regimes. That assumption is wrong. Even a regional carrier may be buying AI-enabled underwriting, claims, CRM, fraud, or service platforms from vendors whose own infrastructure decisions were made elsewhere.
Sovereignty becomes an operational issue the moment AI becomes infrastructure, not just a geopolitical one. This is the same fragmentation IIAI has already mapped across six major regulatory regimes in “Six Regulators, Six Answers.” The insurance layer sits on top of those same questions: NAIC Model Bulletin adoption, the EU AI Act, and Colorado’s shift from its original AI Act to the narrower ADMTA. Those are insurance expressions of the broader sovereignty issues Deloitte identifies.
Autonomy Is Scaling Faster Than the Controls Around It
Deloitte’s most consequential finding for insurance concerns agentic AI, systems that can set goals, reason through multistep tasks, and act with limited human oversight. Financial services firms report 21 percent using agentic AI at least moderately today, a figure expected to reach 71 percent within two years. Set against that trajectory, only 23 percent of financial services companies report having a mature governance model for autonomous agents.
Financial services firms are already moving toward production at a fast clip. Twenty-four percent have pushed 40 percent or more of their AI experiments into live use. Fifty-three percent expect to cross that threshold within three to six months.
The survey is not warning that organizations are experimenting too aggressively. It is showing that they are preparing to industrialize AI faster than they are building the management systems to control it. Deloitte did not need to manufacture tension between those production numbers and the 23 percent governance figure. It is already there.
For insurance specifically, the stakes are sharper than Deloitte’s cross-industry framing suggests, because the workflows in question are high-consequence: underwriting appetite, claims decisions, customer communications. Regulators are still working out how to treat this shift themselves.
The Federal Reserve’s SR 26-2, issued jointly with the OCC and FDIC in April, carves generative and agentic AI out of standard reproducibility requirements in its footnote 3, a tacit acknowledgment that the old governance playbook does not map cleanly onto systems that act rather than merely recommend. OWASP’s Top 10 for Agentic Applications is currently the closest thing to a technical risk framework for these systems, and it exists because the gap Deloitte is describing is not theoretical.
This progression is already visible in insurance distribution technology, where AI tools are moving from recommendation engines toward increasingly autonomous workflows faster than most organizations’ governance models are evolving, a pattern IIAI examined in “Who Makes the Shortlist.” The point there, as here, is organizational maturity, not which vendor is furthest along.
Automation Expectations Are Ahead of Workforce Redesign
Financial services firms expect automation to move fast. Thirty-one percent expect at least 10 percent of jobs to be fully automated within a year, and 81 percent expect that level within three years. Against those expectations, 84 percent of financial services companies have not redesigned jobs around AI capabilities at all.
That gap matters more in insurance than the framing of a generic skills shortage suggests. Entry-level underwriting and claims roles, the kind of routine, judgment-adjacent work Deloitte’s interviewees flag as an early automation target, are also the roles that have historically trained the next generation of senior underwriters and claims professionals.
AI threatens the apprenticeship model before it threatens the profession. The risk is not that AI eliminates the insurance workforce. The risk is that organizations automate the very roles through which that workforce traditionally developed judgment, without building any alternate path for how the next generation acquires it.
These three issues, where AI runs, how AI acts, and who works alongside AI, may look unrelated. They are connected by the same shift. Organizations are moving from AI experimentation into AI integration, and integration forces decisions about infrastructure, control, and people that pilots let them postpone indefinitely.
What This Means for Insurance Leadership
The constraint for insurance was never really whether to adopt AI. It is whether insurance organizations build the process, governance, and workforce structures that turn adoption into operating capability.
For insurance leadership, that shows up as three distinct problems rather than one undifferentiated challenge: regulatory exposure through infrastructure choices made by vendors, operational risk from autonomous systems scaling faster than the controls around them, and talent pipeline erosion as entry-level roles disappear before anyone has redesigned how expertise gets built.
Insurance has largely solved the access problem. The harder challenge, and the one Deloitte’s findings quietly point toward, is turning that access into organizational capability.
The investment decision is increasingly the easy part. The organizational redesign that follows it is where the real work, and this year’s actual strategy question, begins.
Sources
Deloitte AI Institute, “The State of AI in the Enterprise: The Untapped Edge” (2026). https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf
Deloitte AI Institute, “State of AI in the Financial Services Industry” (March 2026). https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/StateofAI-Financial-Services.pdf
Board of Governors of the Federal Reserve System, OCC, and FDIC, Interagency Guidance SR 26-2 (April 17, 2026).
OWASP Top 10 for Agentic Applications, owasp.org.
InsuranceIndustry.ai, “Six Regulators, Six Answers: The AI Insurance Governance Map Nobody’s Drawn” (July 18, 2026). https://insuranceindustry.ai
InsuranceIndustry.ai, “Who Makes the Shortlist: The New Distribution Penalty” (July 22, 2026). https://insuranceindustry.ai/who-makes-the-shortlist-the-new-distribution-penalty/
AI Disclaimer: This content was created with assistance from artificial intelligence technology. While content is based on factual information from the source material, readers should verify all details directly with the respective sources before making business decisions.
