AI in Insurance: Claims Leads Today, Underwriting Is Next
A new report from Sollers Consulting shows AI is now part of everyday operations in insurance, especially in claims. The findings also point to big differences between insurers in how well they scale AI beyond isolated use cases.
The publication, "Beneath the Surface of AI in Insurance," is based on interviews with senior leaders at 35 insurers across Europe and North America, including the UK, Germany, France, Switzerland, the Nordics, the US and Canada.
Where AI Delivers Today
AI is strongest in claims handling, back-office work and customer experience. Motor insurance is a standout area, with automation improving triage, document processing and response times.
Key adoption signals:
- 64% have deployed AI for tasks like document deduplication and email triage.
- Data extraction is the most common use case (applied or in development) at 69% of insurers.
- Chatbots are in use at 60% of companies.
- AI-assisted call centre tools lag at 38% adoption.
Underwriting: The Next Big Shift
Leaders expect AI to reshape underwriting within three to five years. Expect growth in pre-bind risk assessment, submission triage, pricing support and sales enablement.
Interviewed managers also see AI becoming more important in risk prevention and distribution, not just operations.
Customer Impact You Can Measure
For policyholders, the outcomes are clear: faster claims decisions, quicker underwriting, clearer explanations of risk and more individualised pricing. This is where trust and retention are won.
Governance Is the Bottleneck
LLMs are spreading across teams, but oversight often isn't. According to Sollers, 26% of insurers lack a formal AI governance model. Some have rolled out general-purpose AI tools for internal experimentation, which creates duplication and sprawl without coordination.
Piotr Kondratowicz of Sollers warns that insurers without governance for AI will fall behind as the market moves toward automation and predictive decision-making at scale.
What To Do Now: A Practical Playbook
- Stand up AI governance: Define roles, policies and approval gates. Cover model risk management, data privacy, explainability, retention and vendor oversight.
- Simplify your architecture: Centralise shared AI services (OCR, embeddings, model registry), standardise APIs and ensure strong MDM and metadata. Reduce one-off solutions.
- Industrialise data extraction: Scale document and email processing across claims, FNOL, billing and compliance. It's the highest-ROI use case today.
- Target underwriting pilots: Start with submission triage, appetite checks and pricing support for simple risks. Add human review and clear audit trails.
- Advance claims automation: Improve document ingestion, fraud scoring, subrogation detection and reserve support. Track leakage and LAE impact.
- Deploy LLMs with guardrails: Use retrieval-augmented generation, prompt management, redaction for PII and monitoring for hallucinations. Keep humans in-the-loop for decisions.
- Measure outcomes, not activity: Time-to-decision, straight-through processing rate, loss adjustment expense, quote turnaround, hit ratio, NPS and model stability.
- Invest in people: Train underwriters, claims handlers and operations teams on AI use, limits and escalation. Pair SMEs with data science.
Benchmarks and Signals to Track
- Claims: FNOL-to-settlement time, LAE, fraud detect rate, re-open rates.
- Underwriting: submission triage accuracy, bind/hit ratio, quote speed, premium adequacy.
- Operations: document touch time, exception rates, model drift alerts, compliance findings.
Risk and Compliance Resources
For policy and governance alignment, see these reference points:
Skills and Enablement
If your teams need structured upskilling on AI workflows, model risk and practical tools, explore curated learning paths by job role: Complete AI Training - Courses by Job.
Bottom Line
AI is now a core capability in insurance. Claims is delivering value today, and underwriting is next. The winners will be the carriers that pair clear governance and modern data architecture with disciplined, measurable rollout across functions.
Your membership also unlocks:
