29% of Your Workers Are Sabotaging AI. Here's Why.

A new Writer and Workplace Intelligence survey of 2,400 knowledge workers found that 29% of employees admit to actively sabotaging their company’s AI strategy. Among Gen Z, it’s 44%. The same survey caught 75% of executives admitting their AI strategy is “more for show than a meaningful guide to outcomes.”

Read those two numbers in the same breath. Three in four leaders admit their own AI strategy is performative. Three in ten workers are fighting it. Any change consultant will tell you those numbers are related. The workers are not the problem.

The sabotage story broke across the industry press in early April. Fortune framed it as a Gen Z backlash. That read is wrong. This is a strategy failure being absorbed by the workforce, and the workforce is returning it with a note attached.

Quick Verdict

Finding	What the Data Shows
Workers sabotaging company AI	29% admit it. 44% among Gen Z. (Writer, 2,400 workers)
Executives admitting AI strategy is performative	75% call it “more for show than a meaningful guide”
White-collar AI avoidance or rejection	80% are either avoiding or actively rejecting company AI tools
Confidence direction	AI use up 13% in 2025. Worker tech confidence down 18%. (ManpowerGroup)
Executive response	60% of companies plan AI layoffs for non-adopters
Saboteur motivation	26% cite poor AI strategy, not job fear
What this means for you	Your rollout is failing upstream of your people. Fix the strategy before you fire anyone.

What Counts as Sabotage (And Why the Label Is Misleading)

The word “sabotage” is doing a lot of work in the headlines. Here’s what Writer actually caught workers doing.

Entering proprietary data into unapproved public AI tools. Using shadow AI that IT doesn’t sanction. Refusing to touch the approved stack. Tampering with performance reviews that feed into AI evaluations. And, in the most cited example, intentionally producing low-quality output through the company tool so leadership concludes the tool doesn’t work.

Some of that is real sabotage. Some of it is shadow adoption, which is a different problem. Some of it is rational skepticism about a tool the worker has already tested and found wanting. Bundling them into one 29% number makes the story punchier. It also obscures what’s actually happening inside the org.

The behavior that matters most is the last one. Workers deliberately producing bad output so the company kills the rollout. That’s a vote of no confidence in a specific deployment. I wrote about this pattern directly in my piece on the Microsoft Copilot rollback. When a tool gets pushed on a workforce that wasn’t consulted, on workflows that weren’t mapped, against incentives nobody adjusted, the tool gets rejected. Loudly.

The 75% Executive Admission Is the Real Story

Buried under the sabotage headline is a number I can’t stop thinking about. Three out of four executives in the Writer survey agreed that their company’s AI strategy is “more for show than a meaningful guide to outcomes.”

Read that carefully. This is not the workers saying it. This is the executives themselves. The people paid to design and execute the AI strategy saying, in a confidential survey, that the strategy they are currently running is theater.

You cannot run a real AI program on a fake strategy. The workforce can smell it. When leadership announces a “company-wide AI initiative” with a keynote slide, a Slack channel, and a press release, and the middle of the org can’t get an answer to “what am I supposed to stop doing so I have time to learn this,” the workforce correctly concludes the initiative is optics. They respond accordingly.

The 29% sabotage rate is a consequence of the 75% performative-strategy rate.

Why Gen Z Sabotages at 44%

The generational data gets read as “Gen Z is uniquely resistant to AI.” That doesn’t survive five minutes of scrutiny. Gen Z uses AI more than any other generation. They adopted ChatGPT in college. Many built side projects on it. They are not anti-AI.

They are anti-stupid-rollout. And they have less political capital to push back through normal channels.

When a 45-year-old director disagrees with an AI strategy, they schedule a meeting. They raise concerns in a leadership forum. They get a hearing, or at least the appearance of one. When a 24-year-old analyst disagrees, nobody invites them to the meeting. The channels available to them are the informal ones: shadow tool use, low-quality output in the sanctioned tool, vent sessions with peers that become peer-level consensus that the rollout is a joke.

44% is an access problem dressed up as a generational one. Gen Z disagrees by routing around the system because the system doesn’t invite them in.

This is the same dynamic I described in my piece on the AI pilot failure trust gap. Trust doesn’t decay uniformly across the org. It collapses fastest where leaders have the least visibility.

The Confidence Paradox Nobody Is Talking About

The ManpowerGroup 2026 Global Talent Barometer put a stat next to the Writer data that reframes the whole conversation. In 2025, regular AI use among workers rose 13%. In the same period, worker confidence in using technology fell 18%.

Stop and sit with that. The more workers use AI, the less they trust it.

This cuts against the standard executive assumption that more exposure drives more adoption drives more enthusiasm. The data says the opposite. Exposure is driving distrust. Workers are using AI, watching it hallucinate, watching it confidently deliver wrong answers, watching their own jobs get restructured around tools that still fail in obvious ways, and their confidence is dropping.

This is why “more training” is the wrong answer to the adoption gap. The workers who use AI most are not the most enthusiastic. They are the most skeptical. They know what it can and cannot do because they have tested it.

The 80% white-collar avoidance figure reported by Fortune is not ignorance. It’s informed skepticism.

The Layoff Threat Is Making Everything Worse

Here’s where the executive response gets self-defeating. The Writer survey found that 60% of companies plan to lay off employees who won’t adopt AI. Leadership looks at the sabotage number and concludes the fix is coercion.

Look at the saboteur data. 26% cite poor AI strategy, not job fear, as their reason for undermining the rollout. These are workers who looked at the strategy, judged it incoherent, and are voting no with their time. Threatening them with layoffs does not fix the strategy. It tells them leadership’s response to their critique is to fire the critics.

What actually happens when a company fires the AI skeptics? You lose the workers with the most field-tested view of the tool. You keep the workers who go along with whatever gets announced. Your AI program becomes an echo chamber with no early warning system. Six months later, a quiet failure becomes a loud failure because nobody internal was willing to call it earlier.

I wrote about the related dynamic in 95% of AI projects fail. Most AI failures are not technical. They are organizational. Coercion makes the organizational problem worse.

What a Real Rollout Looks Like

If the sabotage rate is a symptom, the treatment is upstream. Here’s the framework I’d use in a VP-level AI program this quarter.

1. Start with a written strategy that names what you’re stopping

Most AI strategies I see are addition-only. “We’re going to add AI to sales, marketing, customer service, finance.” Nobody writes down what the organization is going to stop doing to free up the capacity to learn and deploy. The workforce reads that and correctly concludes the strategy is a press release.

Your strategy document, on one page, needs to answer: What are the three things this team stops doing? Who decides? When? The moment you write that down, the strategy stops being theater.

2. Invite the skeptics into the design

This is the one most programs skip. The workers with the strongest opinions about why the tool will fail are the ones you most need in the room when you design the deployment. Not in a town hall. In the actual working group.

If 29% of your workforce is ready to sabotage the rollout, pick the three most credible skeptics in each department and put them on the design team. Give them veto authority on workflow changes. You will ship a narrower, better-scoped deployment. The sabotage rate collapses because the skeptics helped build the thing.

3. Publish the failure rate

Your workforce already knows the AI tool fails sometimes. Hiding that erodes trust. Publishing it, with an improvement metric attached, earns it back. I covered the adjacent version of this in the AI trust deficit piece. The companies that tell the truth about where their tools fall short build more durable adoption than the ones running hype loops.

A one-page internal dashboard with hallucination rate, workflow completion rate, and error types does more for credibility than any training video.

4. Stop using layoffs as the stick

The 60% of companies planning AI-adoption-linked layoffs are creating the condition they fear. The workers who would have eventually adopted are now adversarial. The workers who are already adversarial have confirmation that leadership is the problem.

Replace the threat with two things: a clear definition of what competence looks like in the AI-augmented version of the job, and a specific path to get there. Six-month window. Named mentor. Measurable milestones. If someone still refuses at that point, the layoff conversation is a performance conversation, not a political one.

5. Align the communication with the strategy

I wrote a full guide on this, and the short version is that your AI communication plan has to answer three questions from every worker’s perspective: What am I supposed to do differently on Monday, what happens if I get it wrong, and what happens if I get it right? If you cannot answer those three questions in 30 seconds, your workforce has the information they need to conclude the rollout is theater.

The Five Questions to Run Before Blaming Your Workers

If the Writer data made your exec team uncomfortable, run these questions in your next leadership meeting before you authorize any coercive action.

1. What percentage of our workforce was consulted before the current AI strategy was finalized? (If under 10%, your sabotage rate is a strategy signal.)
2. For each AI tool we’ve rolled out, can we name three workflows it measurably improves and three it doesn’t? (If you can’t, the workforce has tested it and you haven’t.)
3. What are we explicitly asking people to stop doing to create room for AI adoption? (If nothing, you’re asking for addition without subtraction.)
4. How are we measuring hallucination rate, workflow completion rate, and user-reported errors? (If you aren’t, you have no credibility on “the tool works.”)
5. What’s our path back to trust if a deployment visibly fails? (If it’s “we move on and don’t talk about it,” you’re burning credibility for the next deployment.)

Most exec teams land on one or two. That gap is roughly the gap between your current sabotage rate and zero.

Three Mistakes to Avoid

Treating sabotage as a people problem. The 26% of saboteurs who cite strategy quality, not job fear, are telling you exactly where to look. Fix the strategy, and a large share of the sabotage evaporates on its own. Fire the saboteurs, and you remove the signal without fixing the cause.

Mistaking compliance for adoption. Workers who stop sabotaging because they fear layoffs are not adopting AI. They are performing adoption. The productivity gains never materialize because the workflows are still being worked around, just quietly. The Writer survey caught the loud version. The quiet version is everywhere.

Assuming Gen Z is the problem. 44% is the diagnostic, not the target. A generation that grew up with these tools is resisting your deployment harder than any other cohort. That is information about your deployment, not about them. If the heaviest AI users on your team are the most hostile to your AI program, the program is the problem.

The Takeaway

The Writer survey will be quoted by executives for a year as proof that workers are the blocker. That reading is backwards. 75% of executives calling their own AI strategy performative is a bigger number than 29% of workers admitting to sabotage. The first number causes the second.

Worker confidence is dropping while worker AI usage rises. That is not apathy. That is informed judgment from the people who have tested the tools in the field. Threatening them with layoffs will not fix the underlying strategy. It will produce cosmetic compliance and worse outcomes.

If you lead an AI program, you have a real choice this quarter. Keep running the performative version, escalate enforcement, and watch your best skeptics leave. Or publish a real strategy that names what stops, invites the skeptics in, tells the truth about failure rates, and aligns your communication with what’s actually being asked of people.

The 20% of companies that are going to capture most of the value from AI in 2026 are already making the second choice. I wrote about that quintile in 74% of AI gains go to 20% of firms. They don’t have a lower sabotage rate because their workforce is more compliant. They have a lower sabotage rate because their strategy is real.

Workers are not sabotaging AI. They are returning the strategy to sender. Read the note before you fire the messenger.

---

Related Reading:

• AI Pilot Failure: The Trust Gap Nobody Is Fixing
• AI Strategy Guide: Clear Employee Communication
• The AI Trust Deficit: Copilot and Public Perception
• Microsoft Copilot Rollback: Deploy AI Differently
• 95% of AI Projects Fail. Here’s How to Be in the 5%.