The insurance industry’s AI spending as a share of revenue is projected to triple in 2026, yet just 38 percent of property and casualty insurers are generating value at scale from AI in their core workflows (BCG AI Radar, March 2026). The gap is not technical. It is, at its root, a people problem. And according to Christopher Orr, senior client partner for insurance, reinsurance and retirement at Korn Ferry, the global organizational consulting firm, that people problem is also the industry’s most consequential and most underinvested challenge.
The urgency around AI has entered a new phase. Carriers spent the past several years discovering what the technology could do. Now they are being asked what it is actually worth. “The urgency is moving from AI experimentation to AI economics,” Orr said. “Where can AI change the speed, quality, and cost of a decision?” That shift in framing matters, because so does the way organizations measure what AI delivers.
Orr recalled a conversation with an insurance business leader who had implemented AI on a complex operational approval process that previously consumed three days. The AI shortened it by two. The leader was disappointed, measuring the result against a headline-driven expectation that AI would transform everything. “If someone had come to him and said, hey, I can cut 66.6 percent of your administrative operational time on this aspect, would you write me a check? He’d have bitten your arm off for it,” Orr said. “But because it was compared against a bar of the robots are taking over, it wasn’t.” The ROI was genuine. The benchmark was wrong.
Industry data suggests this pattern is widespread. A Roots Automation survey of insurers in 2025 found that more than 90 percent reported exploring or testing AI, while only 22 percent had fully deployed solutions in production. The explanation is not slow vendors. It is slow adoption cycles driven by the human and organizational conditions inside the carriers themselves.
The traits Korn Ferry is helping insurers identify in AI-forward leaders read like a different kind of job description. Agility. The ability to instill trust. Comfort managing ambiguity and an appetite for experimentation. A strategic mindset oriented toward second-order effects rather than the needs of the moment. Resilience in the face of technology that even its creators do not fully understand. The interpersonal savvy to drive genuine engagement across an organization that may be skeptical, uninformed, or resistant. “You need to cultivate innovation and you’re going to have to manage conflict,” Orr said.
But assembling those traits in individual leaders is, by itself, insufficient. The insight Orr returns to is that AI adoption is a system-level problem, not a talent-level one. “AI doesn’t stall on strategy,” he said. “It stalls when the human system – both people and organizations, incentives, role clarity – don’t move in sync.” The strategy might be clear, the investment approved, the tools deployed, and every employee issued a license. And adoption plateaus anyway. “You need to change goals, decisions, and workflow at the same time as your leaders model actually using AI as a thought partner. When they all move together, it works and scales. If you don’t, it doesn’t.”
BCG’s March 2026 analysis of P&C insurers reaches the same conclusion from a different direction: insurers that fail to generate returns from AI are typically those that drop it into legacy operating models designed for human-led execution, rather than redesigning workflows and roles around what AI can actually do.
When organizations ask who is responsible for making AI adoption work, the answer is often framed around designated roles. Orr answers it differently, using a parallel that reframes the question. When companies first went online, many established digital as a separate business unit, perceived as a competitive threat by existing channels. Over time, digital stopped being a division. It became the medium through which everyone did their job, cloud platforms, SaaS tools, APIs – infrastructure, not a department. “Who owns Excel at a company?” Orr said. “Who owns your intranet? Who owns any tool that you might be using?”
AI is following that same trajectory. The appropriate governance frame is not to designate an owner but to build a culture in which everyone uses it appropriately and visibly. “It’s sort of everyone’s responsibility,” Orr said. “But we’re all testing these things all the time.” The risk of hidden or siloed experimentation is that failures stay hidden too. Organizations that instill trust, role clarity, and open experimentation create the conditions for collective learning. Those that do not find that AI stalls at the edges of the organization, used informally where it works, ignored everywhere else.
If the gap between AI investment and AI returns is a people problem, the major capital allocation decision ahead of the insurance industry is not about compute infrastructure or model selection. Those constraints are real but are being addressed by very capable engineers. “The constraint is about people solving the people sides of these problems,” Orr said. “Are individuals enabled? Has the leadership culture and mindset shifted to support sustained adoption? Have you reimagined the value stream? Are you fully leveraging your data, your technology, your agents? Have you evolved your roles and capabilities? Have you adopted an AI-enabled operating model that allows the more tangible investments in AI to actually be adopted and to provide value?”
The human skills that matter most, Orr argues, are not technical. Technical skills depreciate faster in the AI era than at any previous moment in the industry’s history, because AI dramatically expands what any individual can access. “It’s the human capabilities leveraging the metaphorical library of Alexandria that become exponentially more important,” he said. The ability to make good decisions, to know when an AI output is plausible but wrong, to redirect a confident model, to communicate across organizational levels, to hold context across a long and complex engagement: these are skills that AI cannot replace and cannot compensate for when they are absent.
“You put nitro into my minivan engine and it’s going to blow up,” Orr said. “Put nitro into a stock car and you go real fast. Is your structure able to take the speed?” For carriers asking why their AI investments are not delivering, the answer is usually structural. The technology is not the constraint. The organization is. The carriers that build the right human infrastructure now – the agility, the trust, the role clarity, the cultural permission to experiment visibly – will be the ones that can actually absorb what AI makes possible.
