What Insurance Pros Are Reading: 4 Trends to Act On Now
Over the last 30 days, Best's Review readers zeroed in on four themes: AI data integrity, autonomous vehicles, tech adoption, and outsourced investment management. Here's what matters and how to translate it into action.
AI model poisoning: protect your data, protect your book
Model poisoning corrupts training data so AI systems learn the wrong patterns. For insurers, that can skew pricing, claim triage, fraud detection, and customer interactions without obvious errors on the surface.
- Establish data provenance and approval checkpoints for any external dataset or vendor-supplied corpus.
- Run adversarial evaluations and canary datasets to spot silent drift or unexpected behaviors.
- Separate training, validation, and production pipelines; log lineage for audit.
- Extend vendor due diligence to include model security controls and incident reporting.
- Adopt an AI risk framework such as the NIST AI RMF to formalize policies and testing. NIST AI Risk Management Framework
If your teams need hands-on upskilling on AI use cases and governance, see curated options by role here: Complete AI Training - Courses by Job.
Autonomous vehicles: early costs and liability shifts
As vehicles add sensors and software, repair bills climb even when crash frequency trends down. Liability can move from drivers to manufacturers and tech suppliers, complicating coverage structure and subrogation.
- Refine severity curves for ADAS/AV-equipped fleets; update parts and labor assumptions by make/model.
- Pilot usage-based programs tied to feature activation and real-world performance.
- Train adjusters on diagnostics, calibration, and safe repair standards; formalize OEM collaboration.
- Prepare for more product liability exposure and multi-party claims; align policy language and limits.
Tech progress beyond AI: pick battles that improve combined ratio
Insurers interviewed point to gains from cloud-native cores, API connectivity, and automation alongside AI. The common thread is focus: fewer initiatives, clearer KPIs, faster feedback.
- Choose 2-3 use cases that touch premium or loss ratio (e.g., straight-through SME quotes, FNOL automation, SIU triage).
- Set a 90-day milestone for each: working prototype, baseline metrics, and a go/hold decision.
- Treat data as a product with owners, SLAs, and quality dashboards; add MLOps for any model in production.
- Bake in model risk controls, privacy reviews, and human override-before scale-up.
Outsourcing investment management: specialization is rising
Life/annuity portfolios continue shifting into structured assets, private credit, and other alternatives. That mix demands niche expertise, leading more carriers to outsource mandates or adopt an OCIO model.
- Run a competitive RFP with clear guidelines on fees, reporting cadence, liquidity buckets, and collateral terms.
- Define risk limits by asset type, rating, sector, duration, and concentration; require look-through where feasible.
- Stress test capital and liquidity under downgrades, extension risk, and wider spreads; align with ALM needs.
- Set board-level oversight, SLAs, and termination rights; verify compliance with NAIC designations and RBC treatment.
Best's Review-AM Best's monthly magazine-continues to surface topics that move the market. For company information and research, visit AM Best.
Your membership also unlocks:
