Your people are ready for change. Your operating model isn’t

The real reason transformation doesn’t stick

Most transformations don’t deliver what they promise. The budgets are bigger, the mandates are broader and the urgency is real. Adoption improves briefly, then the organization snaps back to familiar patterns.

Transformation often stalls when systems continue to reward legacy behaviors, even when employees are willing to work differently. The incentives, governance structures, workflows and social norms that determine which behaviors survive under pressure were built for a different operating model. And they are still running, exactly as designed.

Reversion isn’t a breakdown of change. It’s the system working as it should under pressure.

Every major organization has a change management methodology. Most look roughly the same. A model of how people move from unaware to resistant to enrolled to capable, a set of interventions designed to accelerate the adoption curve, and a project team running communications, training and stakeholder engagement in parallel with the “real” transformation work.

The 2025 AI Index Report by Stanford’s Institute for Human-Centered AI found that organizational AI use jumped from 55% to 78% in a single year, with generative AI use more than doubling across business functions in the same period. Yet Gallup found that only about one in 10 employees in AI-adopting organizations strongly agree that AI has transformed how work gets done.

Even when the technology works, the binding constraint shifts to organizational conditions. It’s the problem classic change management was never designed to solve.

The founding insight organizations forgot

The behavioral science that underpins modern change management traces, in part, to Kurt Lewin, the social psychologist who pioneered the study of group dynamics in the 1940s. What most practitioners remember of Lewin is the “unfreeze-change-refreeze” model. What most practitioners have forgotten is the more important finding that came before it.

Lewin argued that it is fruitless to concentrate on changing the behavior of individuals, because the individual in isolation is constrained by group pressures to conform. The focus of change, he insisted, must be at the group level, concentrating on factors such as group norms, roles, interactions and socialization processes. How a person behaves says less about them than about the field they operate in.

What scaled in practice wasn’t Lewin’s field theory. It was a simplified model focused on moving individuals up an adoption curve. That model scaled in part because it was easier to package, teach and operationalize. And the change discipline has been built on that foundation ever since.

In a world where organizations are in constant transformation, those assumptions are now the binding constraint.

Why individual adoption is the wrong unit of organizational change

Classic change management is built on a premise so deeply embedded that most practitioners never examine it: that organizational change is fundamentally an individual phenomenon. Organizations change when enough individuals change. The job of change management is to move individuals through a journey, from unaware to understanding to enrolled to capable, efficiently enough that critical mass tips.

This premise isn’t wrong, but it’s incomplete in a way that makes it dangerous.

Systems define what behavior is feasible, what is rewarded, and what is costly. Over time, behavior tends to align with what the system makes feasible, rewarded and safe. When the system isn’t redesigned alongside the individual, the individual reverts, not because they failed to learn, but because the system is still optimizing for the old behavior.

That optimization happens through incentives, accountability and governance. It also happens through the design of the work itself: how workflows are structured, how teams coordinate and whether AI is embedded directly into the operating model or left optional alongside it. You can change the person. But if the system around them hasn’t changed, the new behavior won’t last.

The AI adoption data makes this failure visible at scale. Massachusetts Institute of Technology research describes a “learning gap,” not in the AI models themselves, but in the organizational systems surrounding them. Workflows that weren’t designed for AI can’t absorb it. Incentive structures that reward existing behaviors undermine new ones. Cross-functional coordination mechanisms that were never robust enough to share insights compound the problem. Organizations are investing heavily in individual capability development while leaving the structural conditions unchanged. The result is widespread experimentation without transformation.

What classic management gets wrong in AI transformation

The failure modes of conventional change management aren’t random. They cluster around three structural blind spots, each of which AI transformation is exposing in sharper relief than any previous wave of change.

Treating trust as an interpersonal problem. Every serious change framework puts leadership trust at the center of the model. The diagnosis is correct and trust is the medium through which change communication travels, but the treatment is wrong. Most change programs respond with more communication, more visibility and more leadership development when the deeper issue is structural misalignment.

AI transformation makes this explicit in a way previous change waves didn’t. A business unit leader told to champion AI-enabled workflow redesign while being measured primarily on short-term output and operational impact faces a genuine structural conflict. The system is asking leaders to optimize for immediate performance and long-term transformation at the same time. In most organizations, the incentives attached to the first are stronger, clearer and more measurable than the incentives attached to the second. The root issue is in governance and accountability. Research on behavioral integrity, the alignment between what leaders say and what they do, consistently shows that perceived integrity gaps are among the strongest predictors of employee disengagement during change. Closing that gap requires changing what leaders are accountable for, not training them to communicate more convincingly.

Confusing information sharing with insight integration. A significant portion of what organizations believe is important for change management is communication, but what they’re missing is the organizational constructs needed to support the desired behavior. Research by Deborah Ancona and David Caldwell on boundary-spanning in teams found that most teams fail not at generating information but at routing it, ensuring that insights generated in one part of the system actually reach and influence the work of another. This plays out clearly in AI transformation. The team running a pilot generates real intelligence about what is working, and that intelligence doesn’t compound because no one is structurally responsible for ensuring it changes how decisions get made elsewhere.

Treating habit formation as an individual discipline problem. The final phase of most change models is about sustaining new behavior, building habits, celebrating wins and reinforcing through recognition and accountability. The assumption is that habits form through personal repetition supported by organizational encouragement.

Author Robert Cialdini’s research on social proof demonstrated that people determine correct behavior primarily by observing what others around them are doing, and that peer influence operates more powerfully than authority influence. People are more likely to be persuaded by a colleague than a superior. An AI workflow adoption that stalls despite repeated leadership communication often takes hold the moment a respected peer starts using it openly, narrating what’s working and pulling others in. Leadership still matters as the force that legitimizes, protects and structurally reinforces the new norms peers then spread through the organization.

Four organizational conditions that make change stick

Classic change frameworks were built for a different era, one where organizations had time to stabilize between shifts. That era is over. What follows are the conditions that actually determine whether behavior change takes hold in organizations that no longer have the luxury of standing still.

At a glance: Four conditions that determine whether a transformation succeeds

1. Leaders who model and architect the change
2. Learning embedded into the organization
3. Shared goals tied to collective outcomes
4. Cross-functional clarity and coordinated action

Leaders who model and architect the change

Leadership accountability has to extend beyond functional results. When it doesn't, leaders end up sponsoring the change publicly while the system keeps rewarding the old behavior privately. Part of the job is redesigning the work itself, how workflows are structured, where decisions get made and whether the new way of working is built into the operating model or left as an option alongside the old one. How performance is measured, and where accountability lands when cross-functional coordination breaks down, determines which one actually wins.

Building learning into the fabric of the organization

People learn through experience, not instruction. When learning stays individual and classroom-based, people return to their roles, and the system pulls them straight back into existing patterns. When teams learn together, working through real friction in real workflows, they develop a shared understanding of how the change affects each role and how those roles fit together.

Shared goals that connect individual success to collective outcomes

Individual goals matter, but they don’t drive adoption on their own. When people are measured purely on functional metrics, they optimize for what’s best for their role, not what’s best for the organization. Without shared goals, that’s exactly what the system produces. When success is defined collectively, people have a reason to pull in the same direction, and that shared commitment is what moves a change from isolated pockets of adoption to something that actually takes hold across the organization.

Cross-functional clarity that drives coordinated action

When roles that have a direct stake in the outcome aren’t part of the process that shapes it, they execute decisions they didn’t make toward goals they didn't set. Coordination breaks down not because people aren’t trying, but because the system never gave them a real stake in the outcome. Every function affected by the change needs a seat at the table, not just the responsibility to execute once the decision is made.

Gallup’s research finds that managers account for 70% of the variance in team engagement, more than culture, compensation or the quality of the change program itself. Managers sit at the intersection of every condition in this section. They’re accountable upward for delivery, accountable downward for their people and surrounded by the same incentivesand peer norms that determine whether anyone on their team actually changes. The conditions surrounding that role matter more than almost anything else in a transformation.

IBM’s global CEO study found that 68% of executives worry that their AI efforts will fail because of insufficient integration with core business activities. They’re diagnosing the problem without having the language for it. What they’re describing is the absence of the conditions above.

A new change management model for organizational transformation

The conditions in the previous section lead somewhere most change programs never go. Classic change management often falls short because it was solving the wrong problem, focusing on execution while leaving the underlying system largely intact.

The old model treats change as something to execute, asking how to move people through a transition and building a project around the answer. The new model treats adaptation as something to design for, asking what the organization needs to look like for new behaviors to emerge naturally, survive competitive pressure and spread without requiring a dedicated program to sustain them. This shift reflects a deeper change in how leaders understand organizations and behavior.

In the old model, the organization is a collection of individuals who need to be guided through change. In the new model, it is a system of conditions that either produces adaptive behavior or suppresses it. The job of a change leader shifts accordingly from communicating, engaging and reinforcing to something harder, building the accountability structures, coordination mechanisms and social environments that make the right behavior the path of least resistance.

AI is accelerating this shift because its pace has made the limitations of the episodic model impossible to absorb. You cannot run a new change program every time a workflow gets reassembled. The old model executes change. The new one builds systems where change can actually take hold.

Stop managing change. Start designing for it

The gap between AI investment and AI-driven organizational adaptation is a systems-design problem. The organizations struggling most are still responding to continuous adaptation pressure with episodic, individual-adoption change models.

Every executive sponsoring an AI transformation right now should be asking whether they have the right questions, not just whether their change program is well executed.

Where will AI accelerate work faster than your organization can socially adapt? Which roles will lose meaning before you create new sources of identity? Where do you rely on informal collaboration that AI will expose as structural gaps? Have you defined leadership expectations that match the speed and ambiguity of AI-driven change? And are you treating change management as communication, or as environment design?

Some are already answering those questions differently, redesigning the systems that produce behavior rather than the people who exhibit it and building the conditions where adopting change becomes the norm.

Most organizations will keep running the same change playbook, adjusting the communications plan, adding training modules, swapping out the framework. The organizations that make progress will redesign the systems that keep pulling people back to the old ways of working.