Every major technology follows the same pattern: excitement, opportunity… and a sudden outbreak of people convinced it will end civilisation. AI is no different, companies are spending heavily on AI, but a growing number of employees are choosing not to use it
Recent research from Writer and Workplace Intelligence found that 29 per cent of employees admit to sabotaging their company’s AI strategy, rising to 44 per cent among Gen Z workers.
That word, sabotage, sounds dramatic. It conjures up images of employees deliberately trying to derail progress. In reality, most AI resistance is likely to be quieter, messier and more human. For instance, ignoring approved tools, using familiar manual processes, feeding work into unofficial systems, or letting poor AI outputs pass through because no one has clarified what good uselooks like.
This should worry leaders, but perhaps not for the obvious reason. The problem is not simply that people are resisting AI. The bigger issue is that many organisations still do not understand why.
A very old reaction to a new technology
The AI saboteur may sound like a modern workplace character, but the pattern is old.
In the early 19th century, the Luddites destroyed textile machinery they believed was threatening their livelihoods, smashing power looms in Nottingham mills and croppping frames in Yorkshire. They were organised enough that the British government deployed more troops against them than it had sent to fight Napoleon. The Captain Swing protests of the 1830s saw agricultural workers break threshing machines, commit arson, and send threatening letters across southern England in protest at mechanisation and the poverty it brought with it.
Even the telephone met resistance. As telegraph and telephone lines spread across rural America in the 1880s and 1890s, some farmers cut wires and toppled poles, convinced the networks were tools of corporate monopoly rather than public progress. In France, early telephone operators faced hostility from café owners who feared the device would keep people at home rather than gathering in public.
Some of this looks absurd with hindsight. Yet it would be too easy to mock the people involved. They were often responding to a genuine change in power, status and economic security.
There is a lesson for AI here. People rarely resist technology in the abstract. They resist what they believe the technology will take from them.
The fear beneath the fear
Most AI conversations in business still focus on tools, policies and productivity. Those things matter, but they miss the emotional centre of the issue.
For many employees, AI is not just another platform to learn. It raises uncomfortable questions about professional identity. If a tool can write the first draft, summarise the meeting, produce the analysis or generate the campaign plan, what exactly is the human being valued for?
That question lands differently depending on role, seniority and career stage. For a junior employee, AI may feel like it is automating the tasks that once helped people learn. For an experienced specialist, it may appear to flatten hard-won expertise into a prompt box. For a manager, it can blur accountability. If AI helped produce the work, who owns the judgment?
This is where resistance starts to make more sense. Some people ignore AI because they do not trust it. Others use unofficial tools because the approved systems feel clunky or poorly explained. Some fear that using AI well may simply prove that their job can be done with fewer people.
Calling all of that sabotage may be satisfying, but it is not especially useful. Leaders need to treat it as evidence.
Adoption fails when AI is applied to people
Much of AI adoption is still being managed like a software rollout. Organisations rush to purchase licences, schedule training and measure usage. Then leaders wonder why behaviour does not change.
The problem is that AI changes work itself. It affects how decisions are made, how quality is judged, how knowledge is shared and how people build credibility. That cannot be solved by giving employees a login and hoping curiosity does the rest.
There is also a credibility gap. Many employees have seen technology programmes introduced with grand promises, only to create more admin, more monitoring or more confusion. When AI arrives with similar rhetoric, scepticism is hardly irrational.
This is important because poor adoption has consequences beyond wasted investment. If employees avoid approved tools, the business loses visibility. If they use unapproved systems, data risk increases. If they rely on weak outputs, quality drops. If they disengage completely, AI becomes another transformation programme that looks impressive to the board’s reporting and disappointing in day-to-day work.
Human value needs to be redefined
The strongest argument for AI adoption is not that machines can produce more. It is so that people can spend more time on the work where human judgment matters. That requires a more honest conversation about value. In many organisations, people have been rewarded for volume, like the number of documents produced, meetings attended, emails answered or tasks completed. AI makes that model look increasingly fragile.
The premium will move towards better judgment. Can someone assess whether an AI output is accurate, useful and appropriate? Can they ask better questions? Can they understand the context behind the data? Can they spot what feels wrong, even when the answer looks perfect?
These are the skills that will separate useful AI adoption from expensive automation theatre.
Leaders should be careful here. Telling people that AI will remove drudgery is unlikely to reassure them if they suspect drudgery is how they prove their worth. The better conversation is about how roles evolve, how people progress and how expertise becomes more visible when routine production is no longer the main measure of contribution.
Sceptics can make AI better
There is a temptation to divide employees into champions and blockers. That is too simplistic. The sceptics may be the people leaders most need to hear from. They are often closest to the messy realities of work, like broken processes, unclear handovers, weak data, duplicated effort, poor customer experience and the quiet workarounds that keep organisations moving.
If those people are resisting AI, instead of dismissing them, ask what they know that the implementation team may have missed.
Good AI adoption should create space for challenge. Employees should be encouraged to test tools, question outputs and flag risks. They should also be involved in deciding where AI genuinely improves work and where it simply adds another layer of complexity.
That does not mean organisations should indulge endless resistance. It means they should distinguish between fear that needs support, scepticism that improves the system and obstruction that needs to be managed.
The leadership test
AI sabotage is a useful phrase because it gets attention. It should not become an excuse for lazy diagnosis.
If employees are bypassing, ignoring or undermining AI, leaders need to look beyond compliance. Have they explained why the technology matters? Have they shown how roles will change? Have they made approved tools easier to use than unofficial ones? Have they created safe ways for people to challenge poor outputs? Have they changed measures of performance so that judgment matters as much as production?
Every major technology changes what people value, what they fear and how they protect their place at work. AI is no different. The so-called saboteur is not just a blocker to progress, but the clearest signal that the business has failed to bring its people with it.
