While organizations are advancing AI initiatives, many AI deployments are stalling because they aren’t properly preparing their teams for ongoing change.

Consider that by the end of 2025, the research firm Gartner predicted that at least 30% of generative AI projects would be abandoned following lackluster proof-of-concept deployments. The research firm found the actual abandonment rate to be much higher, with Gartner’s 2025 follow-on analysis finding that failures reached 50%. A separate Gartner forecast projects that through 2026, organizations will abandon 60% of AI projects due to a lack of AI-ready data.

The failures aren’t only due to poor data readiness, although organizations do have a lot of data hygiene work ahead of them. Ultimately, the high failure rates stem from the “frozen middle” or middle management that becomes paralyzed between executive strategy and frontline execution. It’s managers who understand what leadership wants but cannot or will not translate it into operational reality — not because they are incompetent, but because the incentive structures, workload, and organizational systems they operate in actively discourage the behaviors transformation requires.

The “frozen middle” is frozen in place by the organization itself. “In my experience, roughly 80%-90% of stalled transformation initiatives are not technology problems. They’re authority, ownership, and incentive problems,” leadership consultant at Middle Multiplier Advisory, Jeff Short, said.

It’s People and Process, Not Technology

There’s a lot of data available supporting Short’s assertion. A 2024 BCG survey found that approximately 70% of AI implementation challenges stem from people and process problems, not technology. Yet organizations invert that ratio in their spending. Deloitte CTO Bill Briggs writes that companies devote 93% of their AI investment to data, technology, and infrastructure, reserving only 7% for people-related work: retraining, change management, and role redesign.

The investment asymmetry exacerbates the challenges, as organizations deploy technology budgets against human problems. “The simplest diagnostic question is,” explained Short, is to ask if managers were given better technology tomorrow, would they make different decisions? “If the answer is no, the technology isn’t the bottleneck,” he said.

McKinsey’s research on failed transformation programs also found that 72% of companies identify management behavior and employee resistance as the primary barriers. And it’s happening because management and employees are doing precisely what the organization has trained and incentivized them to do.

First, incentives remain anchored to legacy management systems. Managers advance by protecting headcount, maintaining process stability, and managing upward. No C-suite mandate changes behavior without changes to compensation structures and career advancement criteria as well. BCG’s 2025 analysis on transformation incentives found that at least 50% of short-term incentive pay should be tied to transformation goals for the duration of the program, typically two to three years. Few organizations do this.

Change is often thwarted because organizational accountability is fragmented. Decision frameworks like RACI (responsible, accountable, consulted, informed) are deployed widely but, as McKinsey has noted, clarify role accountability within processes without resolving who actually decides in contested, cross-functional situations. When multiple people believe they own the outcome, decisions stall. BearingPoint’s April 2025 study of more than 300 managers found that 64% of organizations conduct AI training, but only 35% have structured change management programs. Solid evidence that process theater is substituting for actual governance redesign.

Additionally, is transformation too often dictated rather than co-created? When operational teams receive a mandate rather than participate in building a new solution, resistance is often the response.

When diagnosing whether a stalled transformation is a technology challenge versus an authority or incentive issue, Dan Auerbach, executive coach at Dan Auerbach and Associates, explained that he asks midlevel management one question: “When this initiative forces a tradeoff, who decides, and what happens to you if you decide wrong? If the honest answer is that the call routes back upward, or that being wrong is career-limiting while being slow is invisible, then the problem is rarely the technology,” Auerbach said.

The Importance of Alignment

Tighter headcount growth relative to rising workloads, especially for middle management and frontline workers, compounds the challenges of digital transformation and AI change management. Earlier McKinsey research indicates middle managers spend less than a third of their time on their highest-value functions, which are talent and people management, because administrative and individual-contributor work absorbs the rest. BearingPoint’s 2025 research found that middle managers face 50% more work than they can handle, with 41% of their time spent on low- or no-value activities. Adding a transformation mandate to that load, without removing anything from it, produces exactly the passive resistance organizations then blame on culture.

Gallup found that when managers actively support AI adoption, 78% of their direct reports become frequent users. Where manager support is low, that figure drops to 44%. As of February 2026, a Gallup Germany study found that only 21% of employees strongly agree that their manager actively supports their team’s use of AI.

For transformation leaders, the implication is clear: The frozen middle needs to be thawed. Not just through better communication plans or additional workshops, but with changed incentives, real decision authority, and genuine ownership that make the necessary change achievable. In upcoming articles, we will detail how that’s done.