Insurance executives are increasing AI investment while employee confidence moves in the opposite direction. Accenture’s Pulse of Change survey found that 90 percent of 218 senior insurance leaders plan to spend more during 2026, and 85 percent see AI as a route to revenue growth rather than only cost reduction.
At the same time, fewer than 10 percent of companies are redesigning roles around changed processes. Only 40 percent of employees believe training prepared them for new AI responsibilities, and just 20 percent feel they influence how AI affects their work.
The industry risks funding technical scale without building the workforce required to operate it.
AI Moves From Pilots Into Core Processes
Thirty-four percent of insurance companies are rolling out agents across several functions. Nearly a third of businesses are rebuilding complete processes with AI rather than attaching a tool to the old workflow.
Executive use is rising too, placing generative systems closer to strategy and consequential decisions. Leaders remain optimistic even amid discussion of an AI bubble: 47 percent say they would increase spending if it burst, while only a minority expect major reductions.
Confidence does not resolve foundational problems. Thirty-five percent of leaders acknowledge that data strategy and digital capability determine progress, while 54 percent of employees say poor or misleading output reduces productivity and wastes time.
AI cannot support underwriting, claims, service, or distribution reliably when source data is fragmented or quality checks are unclear. Revenue ambitions must include the cost of validation, correction, integration, and oversight.
Jobs Are Changing More Slowly Than Workflows
Regular employee AI use has fallen by ten percentage points since summer 2025, and independent experimentation has declined by 15 points to 39 percent. The drop suggests that access and executive enthusiasm do not automatically produce sustained adoption.
Only five percent of respondents report adjusting job positions to support AI, and 24 percent have continuous learning programs. Employees are expected to supervise new systems without receiving new authority, incentives, time, or career paths.
Security is also weakening. Forty-eight percent of workers feel secure in their roles, down from 59 percent, while 59 percent believe automation makes it harder for young professionals to find work.
Leaders should redesign the role around the new process. That means defining what the system handles, what the employee verifies, how performance is measured, and how expertise develops when entry-level tasks change.
Workforce Trust Is an Operating Requirement
Executives see talent as an accelerator, yet employees are less confident that organisations can manage disruption. Insurance workers also report lower confidence than leaders in the response to economic shocks.
This gap matters because employees encounter the failure modes that strategy decks miss. Claims handlers, underwriters, service teams, and compliance professionals know when outputs ignore context or create extra work. Their participation improves system design and catches risk earlier.
Training should use real cases, including misleading output and escalation. Workers need channels to challenge tools without being judged as resistant and should understand how their feedback changes the workflow.
Metrics must combine revenue, efficiency, quality, customer outcomes, employee workload, and exceptions. A process that appears automated but transfers correction to invisible human labour has not achieved its promised return.
Accenture surveyed 3,650 C-suite leaders across industries and countries, so the insurance findings sit within a wider climate of investment optimism. The source’s central conclusion is still specific: 2026 will favour companies that match technological confidence with commitment to workforce needs.
Insurers know how to price risk over time. They should apply the same discipline to AI transformation. Capital buys platforms and models; durable value depends on people who are prepared, heard, and capable of recognising when the automated decision does not fit the real claim.
