Something strange happened in late 2024 when companies started mandating AI tool usage across their workforces. Employees didn’t revolt. They didn’t picket. They didn’t write angry memos to HR.

They pretended.

According to research cited by BetaNews, one out of every six workers now pretends to use AI at work even when they actually don’t, performing a kind of corporate theater to satisfy executives who check usage dashboards while accomplishing nothing with the technology itself. This quiet deception reveals something important about workplace AI adoption that most change management playbooks miss entirely.

The problem isn’t fear. It isn’t even resistance. It’s disconnection.

The Myth of the Scared Worker

Open any article about AI workplace adoption and you’ll find the same tired narrative: employees fear job loss, so they resist AI, so you need to reassure them. This story is convenient because it positions workers as irrational and management as enlightened. It’s also mostly wrong.

Yes, some workers worry about replacement. That concern is legitimate and not irrational given the breathless headlines about automation coming for every job. But fear of job loss doesn’t explain the engineer who sees AI as technically useful but refuses to incorporate it into their workflow, or the designer who tried AI tools extensively and found them worse than useless, or the manager who mandates AI for their team while quietly avoiding it themselves.

When Section surveyed 5,000 white-collar workers, they found that 40% of non-management employees said AI saves them zero time in an entire week. Not “some time” or “a little time.” Zero. Meanwhile, their bosses remained convinced that AI was transforming productivity. This perception gap isn’t about fear. It’s about reality.

Steve McGarvey, a UX designer quoted in Futurism’s reporting, described his experience: “I can’t count the number of times that I’ve sought a solution for a problem, asked an LLM, and it gave me a solution to an accessibility problem that was completely wrong.”

That’s not fear talking. That’s experience talking.

Why People Actually Resist

The real reasons for AI resistance are less flattering to the technology and more interesting than simple fear.

The tool doesn’t work for them. Different jobs have different relationships with AI capability. Customer service representatives might find AI genuinely helpful for drafting responses. Software architects might find it actively harmful for system design decisions. The marketing coordinator who writes three blog posts a month might save hours. The novelist working on their third book might produce worse prose with AI assistance than without it.

When organizations mandate AI usage uniformly, they ignore this variation completely, treating the technology like email or spreadsheets rather than what it actually is: a powerful but inconsistent collaborator that helps some tasks enormously and others not at all.

The quality bar matters. AI outputs often require the same skill to fix that they would require to create from scratch, which means that for experienced professionals, AI assistance can actually slow things down while degrading output quality in ways that take expertise to notice. The junior employee sees a draft that looks good. The senior employee sees seventeen subtle errors that will take longer to fix than starting fresh.

Identity and craft are real. Writers who spent decades developing voice. Designers who trained their aesthetic sense through thousands of iterations. Engineers who prize elegant solutions over functional ones. These professionals don’t just do their jobs. They are their jobs, and the suggestion that a statistical pattern matcher can replicate their hard-won abilities feels not just wrong but insulting.

This isn’t luddism. It’s pride in craft. And dismissing it as resistance to change misses the point entirely.

The surveillance aspect creeps people out. Many AI implementations come bundled with monitoring: usage tracking, output analysis, productivity dashboards. Workers aren’t paranoid when they notice that the tool designed to “help” them also generates data about their every keystroke. The help comes with a watcher attached, and people reasonably distrust that combination.

What Actually Builds Buy-In

Forget the change management frameworks with their phases and stakeholder matrices. Here’s what actually works.

Start with the skeptics, not the champions. Most organizations do the opposite. They find the enthusiastic early adopters, shower them with attention, and hope enthusiasm spreads. It doesn’t. The skeptics watch the champions succeed and think “that person was already a productivity machine, of course AI helps them.”

Instead, find the people who are most skeptical and work with them directly. If you can address their concerns, you’ve addressed the concerns of everyone less skeptical. And if you can’t address their concerns, maybe their concerns are valid and you should adjust your approach.

A Hacker News user named sevenzero described their situation in a discussion about AI users: “I started to outsource thinking at my job as my company made it very clear that they do not want/cant afford thinking engineers.” This isn’t someone who feared AI. This is someone who adopted it as a survival mechanism in an environment that devalued their thinking. That context matters more than any training program.

Make it genuinely optional, at least at first. Mandates breed resentment. When AI adoption is mandatory, people adopt the minimum viable behavior to satisfy the requirement while mentally checking out. When it’s optional, the people who find it useful become genuine advocates, and their authentic enthusiasm convinces others far more effectively than any executive memo.

The organizations that mandate AI from day one signal distrust. The organizations that offer it as an option and let adoption spread organically signal confidence in both the technology and their people.

Acknowledge when AI is the wrong tool. Nothing builds credibility faster than honest assessment of limitations. When leadership says “we tried AI for this use case and it doesn’t work well, so we’re not pursuing it,” employees learn that the organization evaluates AI honestly rather than pushing it for its own sake.

Research on enterprise AI adoption from Hacker News discussions consistently shows that level-headed assessment beats hype. As one user summarized the consensus: “There’s definite potential, it’s very useful in some specific tasks, but it’s not an all-intelligent panacea.”

Train on actual work, not tools. Most AI training teaches how to use the tool. Few teach how to integrate the tool into existing workflows for existing tasks. The difference matters. An hour learning about ChatGPT features produces nothing. An hour working through how to use AI to accomplish a specific recurring task the employee already does produces immediate value.

The Support That Actually Matters

Training programs typically focus on capabilities. What AI can do. How to write prompts. Which models are best for which tasks. All useful information that misses the actual friction point.

The friction isn’t capability knowledge. It’s integration knowledge. How do I fit this into the work I already do? What changes and what stays the same? How do I judge when AI output is good enough versus when it needs extensive revision?

These questions are role-specific and task-specific and cannot be answered by generic training. A content marketer needs different integration support than a financial analyst. A customer support representative needs different workflow adjustments than a product manager.

The most effective support structure isn’t a training program at all. It’s ongoing access to someone who understands both the AI tools and the specific work being done. Call them coaches, consultants, or just helpful colleagues. What matters is availability when questions arise in the moment of work, not in a classroom separated from context.

This kind of support is expensive and doesn’t scale easily, which is why most organizations opt for the cheaper alternative of recorded training videos that nobody watches twice. But the cheap approach produces the cheap results we see in adoption statistics everywhere.

What Successful Adoption Actually Looks Like

It doesn’t look like universal enthusiasm.

Successful AI adoption looks like pragmatic usage patterns where some people use AI extensively for tasks where it helps, other people use it occasionally for specific narrow purposes, and some people barely use it at all because their work genuinely doesn’t benefit.

This distribution frustrates executives who want hockey-stick usage graphs and uniform adoption curves. But it’s what healthy adoption actually looks like. Not everyone needs to use AI. Not every task benefits from AI. Not every workflow improves with AI inserted into it.

The organizations that understand this distribute AI tools broadly, support experimentation generously, measure outcomes rather than usage, and accept that organic adoption produces uneven patterns reflecting genuine utility rather than compliance theater.

According to BCG’s 2025 research, organizations with formal change management strategies succeed three times more often than those without. But “formal change management” doesn’t mean mandates and dashboards. It means thoughtful attention to the human dynamics of technology adoption.

The worst AI implementations treat people as obstacles to overcome. The best treat them as collaborators in figuring out where AI actually helps.

The Uncomfortable Truth

Here’s what nobody running an AI transformation wants to hear: maybe your resistance problem is actually a value problem.

Not values as in ethics, though that’s part of it. Values as in utility. Maybe the people resisting AI are the ones who’ve actually evaluated it most thoroughly. Maybe their resistance reflects genuine assessment rather than fear. Maybe the tool really doesn’t help them, and their skepticism is data you should be collecting rather than resistance you should be overcoming.

An agency team lead quoted in Piccalilli’s investigation into forced AI usage described their workplace: “They want to be the ‘first AI agency’ and are basically telling us to get on board or you’re not a fit.” That pressure produces compliance without buy-in. It also produces the pretending we discussed at the start: theatrical adoption that satisfies metrics while changing nothing about actual work.

When employees pretend to use AI, they’re sending a message. The message isn’t “we fear change.” The message is “the tool doesn’t help us and you won’t listen.”

Successful AI adoption requires actually listening.

Something To Consider

Organizations spend enormous energy managing AI resistance. They build training programs. They develop communication strategies. They identify champions and address skeptics. All this effort assumes that resistance is a problem to solve.

But resistance is also information. It tells you where the gap exists between what leadership believes and what employees experience. It reveals which use cases actually work and which only look good in demos. It surfaces the concerns that polished vendor presentations never mention.

The organizations that succeed with AI aren’t the ones that overcome resistance most effectively. They’re the ones that distinguish between resistance worth overcoming and resistance worth learning from.

Your people aren’t obstacles to your AI transformation. They’re the transformation itself. The question isn’t how to make them adopt. The question is whether what you’re asking them to adopt is worth adopting.

That question makes executives uncomfortable. It should.