AI will define insurance's future, but coverage is still catching up
AI is reconstructing business models across sectors. In insurance, the bigger issue isn't adoption-it's clarity on how losses get picked up and priced. Andrea Ward, senior vice president, casualty, at CRC Insurance Services, believes the market is underestimating AI as a risk, even as she stays optimistic about its long-term value.
"We don't know what we don't know. AI is evolving day in and day out," Ward said ahead of her panel at the Women in Insurance Los Angeles Summit. Until the industry is clear on claim response, allocation, and exclusions, usage will lag opportunity.
Building coverage for AI-native businesses
The biggest movement today is in E&S. Ward is seeing growth from companies whose core product is AI, not a bolt-on feature. That demand led to a dedicated London facility providing capacity and wording for AI-focused enterprises whose value depends on machine learning.
The exposure isn't just higher frequency or severity-it's new liability paths. Platforms used to connect people. Now, systems create outputs and experiences that can trigger unfamiliar causes of loss. The open question: how does a CGL form apply when the "actor" is an algorithm?
Tech plus judgment-not a single tool that does it all
AI can compress research cycles, surface benchmarks, and widen the view on comparable risks. It can also fuel tougher casualty dynamics by making high verdict data easier to find, which can influence expectations during claims.
Ward's stance is clear: use AI as an assistive tool, not as a final decision-maker. Experience, pattern recognition, and context still decide outcomes.
What brokers should watch next
- Policy intent: Clarify how standard liability forms respond to AI-generated outputs, recommendations, and interactions. Tighten definitions around "occurrence," "professional services," and "media" to avoid gaps.
- Allocation: Expect active debate over which policy responds-CGL, E&O, cyber, media, or tech E&O-especially when multiple policies touch the same incident.
- AI-native facilities: More purpose-built solutions will emerge as carriers get comfortable with loss themes. Track wording evolution and capacity shifts in E&S.
- Data provenance: Push insureds to document datasets, training sources, consent, and licensing. Warranties and representations here can be decisive during claims.
- Model risk controls: Ask about validation, monitoring, human-in-the-loop, and rollback plans. Align submissions with recognized frameworks like the NIST AI Risk Management Framework and NAIC AI Principles.
- Claims posture: Prepare for tougher negotiations where AI-linked outputs, bias allegations, or misinformation are involved. Early expert engagement matters.
- Training: Equip client teams on AI governance, acceptable use, and documentation. Better discipline reduces frequency and improves insurability.
Where coverage clarity is most urgent
- Third-party harm from AI outputs (defamation, IP, misinformation, discrimination).
- Automation errors that drive business interruption or bodily injury/property damage via connected systems.
- Use of unlicensed data or content in training sets (contractual and IP disputes).
- Algorithmic bias claims tied to underwriting, pricing, or claims handling decisions.
People will future-proof insurance
Ward's view is pragmatic: insurance is still a people business. Claims need empathy, judgment, and the memory of "we've seen this before."
Used responsibly, AI can steady E&S by providing more consistent inputs. Seasoned professionals set boundaries, interpret context, and make the final call.
Women in Insurance Los Angeles: keep the conversation moving
Ward will moderate a panel at the Women in Insurance Summit Los Angeles on October 30. Expect practical discussion on applying AI in your role, strategies for career transitions, and building a leadership brand.
If you work in underwriting, broking, or claims and want to level up your AI fluency, explore role-based programs at Complete AI Training.
Quick broker checklist to use this week
- Identify insureds with AI at the core of their product or workflow; segment by use case and model type.
- Map AI loss scenarios to current policies; flag overlaps and gaps.
- Request AI governance artifacts with submissions (data lineage, validation results, oversight structure).
- Pre-agree on incident triage: who is notified first, which experts get engaged, and what evidence is preserved.
Your membership also unlocks:
