Agentic AI hits a tipping point in insurance: 68% of Q4 2025 deployments were gen AI or agentic, Evident finds
Published March 05, 2026, 3:00 p.m. EST | Updated March 05, 2026, 3:00 p.m. EST
AI deployments across insurance surged 87% year-over-year, according to new research from Evident. In Q4 2025, generative and agentic AI represented 68% of all public insurance AI deployments, with agentic AI alone accounting for 21%.
"Insurance is crossing a threshold in AI adoption," said Annabel Ayles, co-founder and co-CEO of Evident. "The rise of agentic can be read as a shift from AI as a productivity tool for individual workers, to AI as an operational system. Right now, AI is starting to run processes rather than just support them."
Where AI is being deployed
In Q4 2025, the most visible use cases were:
- Claims management: 37% of deployments
- Underwriting and pricing: 21% of deployments
- Customer engagement: 21% of deployments
"P&C teams are using AI to automate an ever greater proportion of claims processes, with the most advanced insurers experimenting with end-to-end automation via agentic AI," said Christian Preece, insurance director at Evident. He added that activity is growing in underwriting, pricing, and customer engagement-showing broader AI maturity across the lifecycle.
What agentic AI means in practice
Agentic AI moves beyond point solutions. It plans, takes actions across systems, and closes the loop with outcomes-making true end-to-end automation possible in high-volume, rules-and-data-heavy workflows like claims.
Market scope and sentiment
Evident's analysis spans life, property and casualty, composite, and reinsurance groups in North America and Europe. Complementary research from Digital Insurance shows sentiment lining up with the data: 52% of respondents said generative AI is a top technology trend for 2026 operations, and 24% said the same for agentic AI.
The October-December 2025 survey included agents and brokers (33%), carriers (59%), reinsurers (2%), adjusters (5%), and other insurance professionals (1%).
How carriers and brokers can capitalize this quarter
- Prioritize claims intents: Pick two or three high-volume claim types with clear rules, strong data history, and measurable SLAs. Define the target "auto-resolve" rate and human-in-the-loop thresholds.
- Tighten data foundations: Centralize FNOL, policy, and third-party data feeds. Standardize document intake (OCR + structured extraction) to reduce edge cases.
- Pilot agentic flows: Orchestrate triage → verification → coverage check → liability assessment → payment recommendation. Keep human review for exceptions and high-risk scenarios.
- Measure what matters: Track cycle time, leakage, LAE, reopen rates, NPS/CSAT, and percent of straight-through processing. Set guardrails for fairness and compliance.
- Scale with controls: Stand up model governance, vendor risk checks, audit trails, and prompt/content monitoring. Build rollback plans for model drift and seasonal spikes.
- Extend to underwriting and CX: Use the same playbook for submission ingestion, pre-fill, risk signals, and renewal repricing-plus 24/7 customer agents for policy changes and status updates.
Fast checks before you move to end-to-end
- Clear policy wording and coverage rules the model can interpret
- Reliable integrations for core systems and payments
- Human approval gates for high-value or ambiguous cases
- Bias testing for underwriting and pricing models
- Production monitoring with alerting and sampling for QA
Bottom line
The signal is clear: AI has moved from productivity add-on to operational backbone in insurance. Claims is leading because it's measurable and data-rich, but underwriting, pricing, and customer engagement are catching up. Teams that pick focused workflows, ship controlled pilots, and measure rigorously will bank the gains first.
Looking to upskill your team on practical implementations and workflows? Explore AI for Insurance for hands-on resources and training.
Your membership also unlocks:
