There is a pattern that has repeated itself across every wave of technology adoption in insurance: senior leadership announces a transformation vision, pilot teams demonstrate compelling results, and then execution stalls somewhere in the operational middle of the organization. With AI, this pattern is playing out at scale, and the organizations that recognize it as a structural problem rather than a technology problem are the ones making real progress.
Key takeaways for insurance leaders
- The “frozen middle”: Middle managers control day-to-day operations, and without their engagement and aligned incentives, AI transformation stalls.
- The data reality check: Pilot programs often succeed in controlled environments, but scaling requires unglamorous, non-optional investments in data architecture and governance.
- The ReSource Pro view: Process architecture and workforce readiness are the foundation on which AI value is actually built; organizational readiness is the true differentiator.
Hugh Terry, founder of The Digital Insurer, has spent years working with insurers on AI adoption. His diagnosis of why most initiatives fail to move from pilot to production is pointed and consistently underappreciated: the barrier is organizational readiness, and at the center of that readiness problem sits a layer of management that transformation programs almost never address directly [3].
The layer that controls what actually changes
Terry’s concept of the “frozen middle” describes middle managers who control how work gets done day to day but are, in most transformation efforts, the least engaged stakeholders in the room [3]. Senior leadership sets a digital vision. Front-line staff often embrace new tools. But the managers who structure work, set team priorities, and absorb the friction of process change are typically measured on current performance, not transformation progress. From their vantage point, changing how their teams work carries real professional risk with uncertain personal upside.
This dynamic is not unique to insurance, but it is particularly acute in an industry where middle management layers carry significant institutional knowledge and where performance metrics have historically rewarded stability over adaptation. COOs and heads of operations should recognize this as the central implementation risk in their AI programs, not a secondary people issue to be handled by HR.
Data readiness is not a technology problem either
The second diagnosis Terry offers is equally important. Most insurers believe their data is in better condition than it is. In reality, legacy systems hold data in inconsistent formats, varying quality levels, and fragmented locations [3]. Before meaningful AI can run on that data, the infrastructure needs investment that is neither glamorous nor optional.
This is where many pilot programs produce results that cannot be replicated at enterprise scale. A proof of concept built on a curated dataset in a controlled environment looks like evidence that the technology works. It is actually evidence that the technology works when data conditions are ideal. The transition to production surfaces the gap between pilot conditions and operational reality, and organizations that have not invested in data architecture and governance discover this problem expensively.
From pilot to production requires a different kind of investment
Terry’s framework for bridging the pilot-to-scale gap identifies four requirements beyond the technology itself: change management, process redesign, new skill development, and sustained leadership commitment [3]. Each of these is an organizational capability, not a technology configuration. Organizations that treat AI implementation as primarily a technology deployment are systematically underinvesting in the factors that determine whether the deployment sticks.
The targeted approach Terry describes through his GenAI Explorer program, which specifically engages middle managers as the key lever for organizational change, reflects a correct read of where the unlocking actually needs to happen [3]. If middle management understands the program, has been equipped to lead their teams through it, and has incentives aligned with its success, the organization moves. If they remain bystanders, even the most capable technology deployment will produce uneven adoption and eventually a quiet return to prior habits.
A ReSource Pro perspective
The organizations making durable progress on AI adoption are not necessarily the ones with the most sophisticated models. They are the ones that have addressed the operational preconditions: clean data, redesigned workflows, and management structures aligned with the change they are trying to make. Process architecture and workforce readiness are the foundation on which AI value is actually built. The technology is available to nearly everyone. The organizational readiness to use it at scale is the differentiator.
Source Hugh Terry: AI adoption must avoid ‘Innovation Theater’
Featuring: Hugh Terry, Founder, The Digital Insurer
Publication: The Insurance Lead
Original Publication Date: March 31, 2026
