EY warns insurers short‑staffed for AI
- EY and the Institute of International Finance said in late April that insurers’ risk chiefs are racing to scale AI while still lacking enough skilled oversight staff. - The survey covered 106 insurers, and 80% of CROs put cyber in their top five risks as AI, data and third-party exposure rose together. - That matters because AI is moving from pilots into underwriting, claims and fraud work faster than risk teams can fully staff and govern it.
Insurance companies want AI in the parts of the business that actually move money — underwriting, claims, fraud detection, pricing, customer service. But the people in charge of risk are throwing a small warning flag. EY and the Institute of International Finance said in April that insurers are dealing with a risk environment that is faster, more connected, and more technology-heavy than before, with AI now part of that shift. The headline in the survey was cyber. The more interesting subtext was capacity — insurers want more AI, but they do not have unlimited people who can govern models, monitor data, and keep third-party tools from turning into control failures. ### What actually changed? The new piece of news is the third annual EY/IIF Global Insurance Risk Management Survey, published in April 2026. It drew on 106 organizations across the Americas, EMEIA, and Asia-Pacific, based on work done from November 2025 through January 2026. The report says risk management in insurance is no longer just a defensive function. It is becoming part of strategy and transformation — which is exactly why AI staffing suddenly matters more. (ey.com) ### Why does staffing matter so much here? Because AI in insurance is not one tool. It is a stack of decisions, models, vendors, data pipelines, controls, audits, and human signoffs. If an insurer uses AI to summarize claims files, assist underwriters, flag fraud, or personalize outreach, someone has to test outputs, watch for drift, trace data lineage, document decisions, and answer regulators. The technology can spread across the company faster than the control team can hire. (ey.com) That is the basic squeeze. ### Where is AI spreading first? A lot of it is landing in underwriting and claims. EY’s insurance work says 74% of firms identified predictive analytics as a key area for underwriting and claims, and recent GenAI adoption has been shifting from back-office experiments toward more front-office uses. Claims is especially attractive because the return can show up quickly — faster file review, lower handling costs, better triage, and less manual paperwork. But those are exactly the workflows where bad outputs can create customer harm or compliance trouble. (ey.com) ### Why are CROs so focused on cyber then? Because cyber is still the biggest visible risk bucket. In the 2026 survey, 80% of insurance CROs ranked cyber among their top five enterprise risks, and 30% called it the top threat. AI does not replace that concern. It compounds it. More models mean more data exposure, more vendor dependence, and more operational complexity. So the staffing problem is not “AI instead of cyber.” It is “AI on top of cyber.” (ey.com) ### What kind of talent is missing? Not just data scientists. Insurers need model-risk people, governance leads, data stewards, compliance specialists, cyber staff, and business operators who understand how AI changes a claims or underwriting workflow. EY’s framing is that next-generation risk management needs high-quality data, advanced technology, and a digitally skilled workforce. That last part is the bottleneck. Buying software is easier than building a team that can challenge it. (ey.com) ### Does this change buying decisions? Yes — probably more than the headline suggests. If internal teams are thin, insurers will favor tools that come with monitoring, documentation, audit trails, and centralized controls. That is why EY keeps pushing “AI ops,” governance frameworks, and control-tower style oversight. In plain English, insurers do not just want smarter models. They want models that are easier to supervise with the staff they actually have. (ey.com) ### So is this a slowdown for AI? Not really. It looks more like a filter. AI adoption is still accelerating, and insurers are still pushing it into core workflows. But the winners may be the carriers that treat talent as infrastructure — training underwriters, claims teams, fraud units, and risk staff together instead of bolting governance on later. The catch is simple: in insurance, an AI strategy is only as real as the people available to control it. (ey.com) ### Bottom line? The EY warning is less “insurers are behind on AI” and more “insurers are underbuilt for governing AI at scale.” That is a subtler problem, but a more important one. In this industry, the hard part is not getting a model into production. It is keeping it trustworthy after it gets there. (iif.com) (ey.com)
