Organizations face an urgent challenge in managing change. Leaders are convinced that artificial intelligence will transform their businesses, but the people needed to carry out that transformation have stopped trying, or so it seems. According to McKinsey’s Superagency in the Workplace report, employees are already using generative AI three times more than their leaders realize. However, only 1% of companies say AI is fully integrated into the way work is done. The workers are moving. Organizations are not. Much of that activity, as we will see, occurs entirely outside approved systems: less a sign of resistance than a sign of unmet need.
We’ve seen this pattern in industries on both sides: Tomer as chief customer officer at WalkMe, on the front lines of digital adoption, and Jenny as an executive coach and organizational change consultant. What looks like resistance is often a rational response to a system that changed from above without attracting people. Leaders who close the gap don’t start by tightening control. They start by rebooting the system. Here are three strategies to do it.
First, understand why employees resist
When employees disengage from AI, we call it resistance. WalkMe’s State of Digital Adoption survey tells a more nuanced story. A 52-point trust gap separates executives and employees: 61% of executives trust AI to make complex decisions; only 9% of workers do so. According to McKinsey’s State of AI survey, while 88% of organizations use AI in at least one business function, nearly two-thirds are still piloting rather than scaling. Leaders believe the tools are working. Employees live a different reality. These are not two sides of the same conversation. They are two different belief systems.
Beneath that chasm are five recognizable patterns:
- “I don’t know what I’m supposed to do with that.” Gallup research links resistance directly to loss of control and unclear expectations.
- “I tried it and wasted my time.” More than 80% of AI projects fail, with the main causes being lack of skills, data preparation, and poor workflow integration.
- “I’m afraid of what this means for my job.”—FOBO (Fear of Becoming Obsolete) is real. Workers see the design headlines and connect the dots.
- “No one showed me how.” Most organizations provide one-off or outdated training without the structured learning paths that people need on a day-to-day basis.
- “I’m good at my job. I don’t need this.”—This is the artisan identity, and it’s more of an advantage than a hindrance. As Jenny has explored in her research on healthy friction, the tension between experience and new tools, when channeled well, becomes an engine of growth, not a barrier.
These are not obstacles that have to be overcome. They are signs to read.
1. Give people a clear destiny, not just a directive
Across industries, we see the same pattern repeating itself. An enterprise AI platform is launched with fanfare: executives send a memo, IT licenses are obtained, and a training webinar is posted on the intranet. And then, there are not many changes. Research consistently finds that most AI initiatives fail to achieve the expected results. The employees do not rebel. They simply don’t know what “using AI” means for their role. The directive is clear. Destiny is not.
One WalkMe customer faced exactly this pattern. Employees had access to multiple AI tools, but they wrote vague prompts, got inconsistent results, and gave up. To solve this challenge, reduce cognitive load, and reinforce desired behaviors, the client’s digital adoption team created a library of personalized messages organized by function and use case (over a thousand templates) that gave each person a concrete starting point. An engineer knew exactly which message to use to review the code. A marketer had templates ready for campaign summaries. After a month, abandonment decreased and thousands of interactions were recorded. Same tools. The same people. Different destination. That result was the result of a defined business objective. The goal was not to “increase AI adoption,” but to “reduce first draft time in half for every feature that impacts customer work.” Measurable. Property. Linked to results that were already important to the business.
Instead of “use AI more,” try: “By next quarter, the first draft of any client deliverable should take half as long, and here’s exactly how.” That is a destiny.
Questions to direct your team:
- Have you defined what AI success looks like for each role?
- Does each employee have a specific use case to start with?
- Is your destination specific enough that someone can confirm that you have reached it?
- What does “using AI well” look like in your team’s daily workflow?
2. Connect AI adoption to what people already care about
People are not moved by logic or commands. They move toward what they find rewarding, affirming of their identity, and safe. This is precisely where most AI implementations fail: treating adoption as a compliance issue rather than a human issue.
What people really want from their jobs doesn’t change because AI comes into it: to feel competent, not exposed; doing visible work, not invisible; do work that matters, not work that anything can do. AI adoptions succeed when it is framed around those needs rather than a mandate, a dynamic that McKinsey has linked to self-determination theory, which holds that employees are autonomously motivated when their needs for competence, autonomy, and relatedness are met. The rethink is simple but momentous: Stop asking employees to “embrace AI” and start asking them what kind of professionals they want to become. An expert analyst who sees AI as a threat to their expertise will resist. That same analyst, invited to become the person who produces better insights faster, leans in. The same tool. Different frame.
One organization Tomer works with evolved its digital adoption team from SaaS enablement to a team focused on helping build AI fluency across the enterprise: human AI experience design, AI-enabled workflows, and role-based rapid selection. The team’s framework moved from “we have to use AI” to “understanding AI and driving AI fluency is a great opportunity to make a significant impact.”
The expanded scope gave the team a different kind of work: less repetitive, less stress caused by friction, and more room to focus on higher-value work. That is a change of identity and it spreads. What made it durable was that business, IT, and learning leaders operated from a shared definition of success. Each function had a part (infrastructure, competition, results) and together they could see the whole picture.
Questions to motivate your team:
- What does your team already care about, and how does AI help them do more of it?
- Have you created visible career markers for AI fluency, or is adoption invisible and unrewarding?
- Have you invited employees to publicly commit to a specific AI use case? Small commitments that are visible tend to last.
- Are you viewing AI as a threat to your abilities or as an amplifier?
- Is there psychological safety to experiment, fail and try again, or just pressure to perform?
3.Make correct behavior easier than incorrect behavior
Nearly half of employees admit to using AI tools without employer approval, and many share sensitive data in the process. The instinct is to take drastic measures. But that misinterprets what is happening. Workers are not rebelling against governance: they are following the path of least resistance. Approved tools are harder to access, less integrated, or simply unknown.
One global professional services firm Tomer worked with had a persistent bottleneck: Identifying the right cost center for customer engagement required manual searches across dozens of options. They incorporated AI directly into that step: what required multiple searches became a single click, in the same place where employees already worked. The adoption was immediate; not because the behavior has changed, but because it was not necessary. Don’t ask people to adopt AI. Make AI part of how they already work.
Questions to shape the path:
- Where could AI be integrated directly into existing workflows?
- What makes bypassing approved AI easier right now than using it?
- What small changes (a template, a shortcut, a default message) could make the correct behavior seem automatic?
- How can shadow AI be treated as a diagnosis rather than a disciplinary issue?
Take off, together
Closing the AI adoption gap doesn’t require better tools or stronger mandates. It requires directing people toward a clear destination, connecting change to what they already care about, and creating an environment in which the right behavior is also the easiest.
Your people are not waiting to be pressured. They are waiting to be guided. Commands drive behavior. Meaning moves people.