Commercial Insurance Embraces AI While Preserving Human Oversight
AI is compressing cycle times and sharpening risk selection, but underwriting judgment still sets the guardrails.
AI is remaking how carriers evaluate, price, and manage risk. Lockton's research points to a simple truth: more data, cheaper compute, and rising pressure on cost and speed are pushing P&C teams to rethink the work.
Data inputs have multiplied through IoT sensors, telematics, and aerial imagery. Cloud-scale processing has made analysis faster. At the same time, losses, talent gaps, and client expectations for transparency are forcing new operating models.
Three Adoption Models Are Taking Hold
- AI Innovators: High automation. Binding quotes in seconds for high-volume, low-complexity lines like personal auto and homeowners.
- Measured Adopters: Targeted use in middle-market and specialty. AI handles ingestion and risk scoring so underwriters can concentrate on exceptions and complex accounts.
- Cautious Followers: Decision support only. Systems flag anomalies and offer pricing benchmarks while humans make final calls.
Where Value Shows Up
- Carriers: Faster quotes, tighter segmentation, lower expense ratios, and broader appetite. Pattern detection across large datasets supports more precise pricing and healthier portfolios.
- Brokers: Shift from transactional to advisory. Benchmarking and predictive tools help structure programs, pressure-test limits, and explain trade-offs.
- Buyers: Continuous risk tracking, better predictive modeling, and sharper exposure identification. Benefits reach beyond premium to resource allocation and operational controls.
Risks That Demand Guardrails
Bias is the headline risk. Flawed data or weak assumptions can push discriminatory outcomes and erode trust. The black box concern is real: without clarity on sources and logic, it's hard to explain or defend decisions.
Regulators are raising the bar on privacy, fairness, and explainability. See the NIST AI Risk Management Framework and the NAIC Model Bulletin on AI. Overreliance on biased or weak models can lead to underinsurance or poor capital allocation-and potential negligence claims for leadership.
Keep Humans in the Loop
Technology scales judgment; it doesn't replace it. Underwriters still need to ask why a model is confident, what data it learned from, and where it breaks.
Core skills stay relevant-underwriting, actuarial, analytics-but the toolset changes. Teams need fluency in data lineage, model logic, and digital workflows. Brokers and buyers should get comfortable with algorithmic assumptions and the quality of inputs driving recommendations.
90-Day Action Plan
- Map your AI footprint: List use cases by line, segment, and decision criticality. Define thresholds for human review and override.
- Stand up model governance: Track data sources, approvals, and monitoring. Run bias and performance tests by segment; document changes and drift.
- Tighten data controls: Set retention, de-identification, and access rules. Add privacy and security requirements to vendor DPAs.
- Clarify accountability: Name an owner for each model. Require clear escalation paths and an appeal process for adverse decisions.
- Upgrade client comms: Explain where AI is used, how decisions are reviewed, and options for human reconsideration.
- Tune contracts: Add audit rights, model change notifications, and data rights. Push vendors for transparency on inputs and features.
- Track the right KPIs: Quote time, hit ratio, loss ratio drift, model override rate, exception cycle time, and adverse change events.
- Invest in skills: Train teams on model basics, data quality, and prompt discipline. For structured upskilling by role, see practical AI upskilling by job function.
Bottom Line
AI can speed decisions and sharpen pricing, but trust comes from oversight. Pair data and models with seasoned judgment, document the why behind each decision, and keep a human on the hook where it matters most.
Your membership also unlocks:
