U.S. P&C Insurers Pour Billions into AI as Core Operations Get Rebuilt
AI has moved from pilot projects to the backbone of daily work across underwriting, claims, service, and fraud. Carriers are chasing efficiency, lower expense ratios, faster cycle times, and better customer outcomes. The spending wave that began in 2022-2023 has only intensified in 2025, and it's changing how insurers compete.
What's actually different now
Insurers are processing real-time data at scale: weather feeds, telematics, drone and satellite imagery-far beyond static datasets and manual review. The big shift is GenAI. Fine-tuned large language models now handle cognitive work such as summarizing medical notes, drafting correspondence, and standardizing documentation.
Reported results are material: up to 20% lower claims operating costs and productivity lifts above 30% for teams using AI assistants. Confidence is high too-an industry survey in late 2023 found 89% of insurance investment professionals see the benefits outweighing the risks.
Where carriers are applying AI right now
- Underwriting: Risk segmentation from dynamic data, submission triage, appetite checks, loss cost modeling, and straight-through processing for simple risks.
- Claims: First notice triage, document ingestion, medical note summarization, liability guidance, and AI vision for damage assessments.
- Service: AI assistants for frontline reps, intelligent routing, intent detection, and automated follow-ups.
- Fraud/SIU: Anomaly detection, network analysis, and case scoring that flags patterns humans miss.
- Marketing & comms: Drafting letters, FAQs, and campaigns with strict brand, compliance, and audit controls.
Ecosystem effects and the widening gap
AI-focused insurtechs dominated Q3 2025 fundraising, capturing 74.8% of all funding across 49 deals. P&C-focused insurtechs saw a 90.5% jump to $690.28 million. Early, aggressive adopters are pulling ahead.
Carriers like Allstate (NYSE: ALL), Travelers (NYSE: TRV), Nationwide, and USAA are viewed as AI leaders based on spend and talent. Vendors such as Gradient AI (underwriting) and Tractable (visual damage assessment) are seeing strong demand. Firms like Lemonade (NYSE: LMND) set claims speed expectations, while Root (NASDAQ: ROOT) pushes behavior-based pricing. Even general AI platforms are being adapted for industry-grade use cases.
Proof points from insurers
Aviva reported a 30% improvement in routing accuracy, a 65% drop in customer complaints, and savings near £100 million after deploying AI across service workflows. CNP Assurances lifted automatic acceptance for health questionnaires by 5%, surpassing 80%.
Key risks-and how to manage them
- Privacy and data protection: Implement strict data minimization, PII controls, and retention policies.
- Fairness and bias: Use bias testing, representative training data, and model governance with clear accountability. Review NIST's AI Risk Management Framework and NAIC AI Principles.
- Explainability and model risk: Keep documentation, lineage, feature catalogs, and challenger models. Establish human-in-the-loop for high-impact decisions.
- Third-party risk: Contract for audit rights, data residency, and incident SLAs; validate vendors' evaluation metrics against your loss drivers.
- Operational controls: Add prompt controls, red teaming, toxicity filters, and watermarking for GenAI outputs; log everything.
Budgets and the road ahead
Experts expect AI to climb from roughly 8% of IT budgets to near 20% within three to five years. Near-term use grows across legal review, policy language analysis, and churn prediction. Longer-term, expect more dynamic products that adjust to real-time behavior and simpler claims handled end-to-end without human touch-backed by stronger model governance.
90-day action plan for P&C leaders
- Choose 2-3 use cases tied to expense ratio, LAE, or NPS (e.g., claims doc intake, FNOL triage, submission triage).
- Form a tiger team: product owner, underwriting/claims SME, data engineer, ML engineer, compliance, and security.
- Data readiness: map sources (policy, claims, billing, third-party), fix quality issues, and define golden labels for evaluation.
- Architecture: decide build vs. buy; pick an LLM access pattern (API, hosted, or on-prem), vector store, and observability stack.
- Guardrails: set prompt policies, PHI/PII handling, retrieval boundaries, and human approval for sensitive outputs.
- Pilot in production-lite: 4-6 weeks, measurable KPIs (cycle time, touch time, leakage, accuracy, CSAT), and A/B testing.
- Upskill teams: train adjusters, underwriters, and service reps on new workflows and quality checks. If you need role-based paths, see AI upskilling by job function.
- Vendor diligence: evaluate security posture, fine-tuning options, content filters, and cost per outcome (not just per token).
- Scale with governance: implement model risk management, drift monitoring, and periodic fairness reviews.
What to watch next
- Fresh rounds of AI spend and carrier-vendor partnerships.
- New products with dynamic pricing and near-instant claims for low-severity losses.
- Regulatory updates on fairness, transparency, and audit expectations.
AI now sits at the center of P&C strategy. The carriers that move decisively are widening the gap with faster claims, smarter pricing, and leaner operations. Those results compound.
This content is for informational purposes and reflects analysis of current AI developments.
Your membership also unlocks:
