A company buys everyone a licence, runs a workshop on prompting, sends people back to the same jobs with the same deadlines and the same targets, and finds six months later that nothing has changed. The internal explanation is that staff are resistant to change. On the evidence, that explanation is almost always wrong. Employee resistance to AI exists, but it is a minority effect, it is rarely about the technology, and where it appears it is usually a reasonable response to something the business did or failed to say. The likelier reason a rollout produces nothing is not that people refused to change how they work. It is that nobody changed the work.
This picks up where your AI feature works and nobody is using it left off. That post was about the gap between a working tool and an adopted one. This one is about the accusation that gets made when the gap appears, and why it sends the diagnosis in the wrong direction.
The data does not support the resistance story
Gallup has asked American employees the same question every quarter since 2023, which makes it the cleanest time series available. In the second quarter of 2026, 52% of US employees said they use AI in their role, 30% used it a few times a week or more, and 15% used it daily. Total use was 21% three years earlier. The sample was 22,573 employed adults, fielded between 6 and 20 May 2026, with a margin of error of roughly one percentage point.
The more revealing figure sits next to it. In the same study, 47% of employees said their organisation had integrated AI tools. Individual use is running ahead of what employers have actually put in place. Whatever is happening in the average workplace, it is not a workforce refusing a technology their employer is pushing on them.
BCG's fourth annual AI at Work survey, published on 3 June 2026, found the same from another angle: 74% of frontline white-collar employees are now regular AI users, up more than 20 percentage points over two years. It covered 11,749 workers across 14 markets, including 503 in South Africa, and South African respondents reported regular use above the global average, alongside India, Brazil and the Middle East.
McKinsey put a number on how badly this gets misread. Its Superagency in the Workplace report, published in January 2025, found that C-suite respondents estimated 4% of their employees used generative AI for at least 30% of their daily work. Employees self-reported 13%, about three times as many. In the same survey, leaders were more than twice as likely to name employee readiness as a barrier to adoption as they were to name their own role.
Two caveats belong in the same breath. The fieldwork was October and November 2024, so it is dated. And the population is enterprise: every US C-suite respondent worked at a company turning over at least a billion dollars. As we have argued before, most AI advice is written for companies that don't look like yours, and that applies to the research underneath the advice. What transfers is the direction: leaders underestimate how much their people already use AI, and over-index on employee readiness as the obstacle. The 4% against 13% split itself does not transfer, because it was not measured in a business doing R15 to R50 million.
Where employee resistance to AI is real, and what it is about
Something real does sit underneath, and it is more specific than fear of technology.
BCG's clearest finding is about time. Of frontline employees who use AI regularly, 42% report saving at least a full working day a week. Of those reporting savings, 66% say they receive limited or no guidance on what to do with the recovered time, and more than half do not redirect it into more strategic work.
Read that from the desk rather than the boardroom. You have found eight hours a week. Nobody has told you what those hours are for. Two inferences are available: the hours become better work, or the hours become more work. Without a stated answer, the second is the safer bet, and acting on it is not resistance. It is arithmetic.
The same logic applies to the knowledge itself. Research published in Harvard Business Review in June 2026 found that employees who develop useful AI methods privately often decline to share them, and that organisational trust and psychological safety predicted disclosure more strongly than formal AI policy or sanctioned tooling did. The reasons given were being judged less capable, being handed more work, or making one's own role look easier to replace. The earliest large signal pointed the same way: Microsoft and LinkedIn's 2024 Work Trend Index, covering 31,000 people across 31 countries, found 53% of those using AI on their most important tasks worried it made them look replaceable.
Job fear is present without being the top concern. In McKinsey's survey, employees ranked cybersecurity (51%) and inaccuracy (50%) ahead of workforce displacement (35%).
The local picture is also not the American one. In the South African figures released with BCG's 2026 survey, about 20% of respondents said they were concerned about losing their job to AI, against 36% globally, and 78% of South African frontline workers reported higher job satisfaction from using it, against a global average of 57%. South Africa also ranked second of the fourteen markets for time saved, behind India, with around two-thirds of South African frontline employees reporting they save at least a full working day a week. Importing US workplace anxiety into a Cape Town or Johannesburg operation and treating it as your diagnosis would be a mistake. The likelier problem here is plainer: the time appeared and nobody said what it was for.
The uncomfortable number, in both directions
Gallup asked laid-off American workers, in their own words, why they were let go. One percent named AI or automation. The most common answers were organisational restructuring (15%), budget or cost cutting (11%) and economic conditions (11%). Gallup states the caveat itself: restructuring and cost cutting can carry AI's influence without anyone being told, so 1% probably understates the indirect effect.
The same analysis, published 17 June 2026, found something more useful. Among laid-off workers, 62% were AI non-users, against 50% of those still employed, and 22% were frequent users against 28% of the employed. The pattern held after accounting for age, education, industry and time since the layoff. In technology specifically, Gallup reported that workers using AI less than monthly were about three times as likely to have been laid off as those using it at least monthly.
That cuts uncomfortably both ways, which is why it is worth saying. To the employee who suspects that learning the tool hastens their own replacement: on the measured evidence the reverse is closer to true, and never touching it looks like the riskier position. To the owner: that is a stronger argument than reassurance, and it only works if you say it plainly and then do not contradict it three months later.
Ask the person doing the job, not only the person who owns the problem
McKinsey found that fewer than half of C-suite leaders, 48%, said they would involve non-technical employees in the early stages of building an AI tool, meaning ideation and requirement gathering. That is most of the failure in one number. The people who know which step of the Friday report is slow, which customer writes their order numbers backwards, and which approval nobody has needed since 2023, are absent when the tool gets specified, and are then held responsible when it does not fit their work.
Gallup's April 2026 analysis of adopters and holdouts found the two strongest separators were whether AI fitted naturally into an existing workflow and whether managers actively supported its use. Neither is discoverable from an executive interview. Both are obvious inside ten minutes of watching someone do the job.
This is the same discipline as the five tests in AI workflows: how to find the ones actually worth automating: does it happen often, is it words-shaped, could a competent new starter do it from written instructions, is being wrong cheap, and is the data reachable. Three of those five can only be answered honestly by the person who does the task. Making the result land inside the systems they already open every morning is AI integration work, and that is usually where most of the effort goes.
BCG found where the difference actually comes from: an explicit strategy lifted AI's measured business impact by 25 percentage points, while better access to tools lifted it by five. The answer to a stalled rollout is rarely a better tool.
What to do instead of running another workshop
- Write down what the saved time is for, before anyone starts. Growth, faster turnaround, fewer weekend hours, taking on work you currently turn away. Any of those is a legitimate answer. Silence is not, because silence gets filled with the worst available guess.
- Ask the person who does the job which step they would delete. Not what AI could do for them. Which part of Thursday they would remove if allowed to. That question produces a better project list than any executive workshop.
- Answer the headcount question once, plainly, in writing. If roles will change and nobody will be let go, say so. If you cannot promise that, do not imply it. People calibrate on what happens next, not on the tone of the announcement.
- Pick a first project where nobody in the room is the subject. Supplier invoice extraction, routing inbound enquiries, drafting standard replies. Boring, high-volume, low-stakes work builds the trust you need for anything closer to the bone.
- Put it where the work already happens. The biggest single cause of a good tool going unused is asking people to go somewhere new.
- Measure behaviour, not licences. Seats provisioned is the easy thing to count and the wrong one. Write down before launch which behaviour should change and how you will see it.
None of that is change-management theatre. It is a handful of cheap, specific decisions that mostly involve talking to the people who do the work before deciding what to build for them. The businesses getting returns from AI are not the ones with the most compliant staff. They are the ones who worked out what to automate by asking, then said out loud what the recovered hours were for. If you want help choosing which task to start with, and what to tell your team about it, that is the first conversation in any sensible AI consulting engagement.
FAQ
Are employees actually resistant to AI? Mostly no. Gallup's Q2 2026 study found 52% of US employees use AI in their role and 30% use it a few times a week or more, while only 47% said their organisation had integrated AI tools at all. Individual adoption is running ahead of employer provision, which is the opposite of the resistance picture.
Why does nothing change after we roll out AI tools? Usually because the tool was added to the work rather than the work being changed. Gallup's adopter research points to two conditions: whether AI fits an existing workflow, and whether managers actively support its use. A licence and a prompting workshop address neither.
Do employees hide their AI use from managers? Some do. Research in Harvard Business Review in June 2026 found employees often withhold AI methods they have developed, and that trust and psychological safety predicted disclosure better than formal policy did. Microsoft and LinkedIn's 2024 survey of 31,000 workers found 53% of those using AI on important tasks worried it made them look replaceable.
Is the fear of AI replacing jobs justified? Only 1% of laid-off US workers in Gallup's Q1 2026 data named AI or automation as the primary reason, though Gallup notes restructuring and cost cutting may carry AI's influence indirectly. More striking: 62% of laid-off workers were AI non-users, against 50% of those still employed, which suggests avoiding the tools is the riskier position.
Are South African employees more anxious about AI than employees elsewhere? Less so, on the available evidence. In the South African figures released with BCG's 2026 AI at Work survey, around 20% of respondents were concerned about losing their job to AI, against 36% globally, and South African frontline workers reported higher job satisfaction from AI use than the global average.
Who should we ask when deciding what to automate? The people doing the task, not only the executives who own the outcome. McKinsey found fewer than half of C-suite leaders would involve non-technical employees in ideation and requirement gathering, which is where fit gets decided. Three of the five useful tests for whether a job is worth automating can only be answered by whoever currently does it.
What is the single most useful thing to do before starting? State in writing what the recovered time is for. BCG found 42% of regular frontline AI users save at least a full working day a week, and 66% of those receive limited or no guidance on what to do with it. That silence, rather than the technology, is what people are responding to.
References
- Gallup: Organizational AI Adoption Jumps Six Points (Q2 2026 workforce study, 22,573 employed US adults, fielded 6 to 20 May 2026). https://www.gallup.com/workplace/712736/organizational-adoption-jumps-six-points.aspx
- Gallup: U.S. Workers Continue to Report Downsizing (17 June 2026; reasons for layoff in workers' own words, AI use among laid-off versus employed workers). https://www.gallup.com/workplace/711287/workers-continue-report-downsizing.aspx
- Gallup: AI in the Workplace: What Separates Adopters and Holdouts (April 2026; workflow fit and manager support as the leading predictors of use). https://www.gallup.com/workplace/704252/workplace-separates-adopters-holdouts.aspx
- Boston Consulting Group: AI Is Reshaping Jobs Faster Than Companies Are Reshaping Work (press release, 3 June 2026; fourth annual AI at Work survey, 11,749 workers across 14 markets). https://www.bcg.com/press/3june2026-ai-reshaping-jobs-faster-than-companies-reshaping-work
- Boston Consulting Group: AI at Work: Why Strategy Matters More Than Tools (2026; time saved, guidance on recovered time, and the impact of strategic clarity versus tool access). https://www.bcg.com/publications/2026/ai-at-work-why-strategy-matters-more-than-tools
- Zawya: South Africa's AI boom is already changing jobs (June 2026; the South African cut of BCG's AI at Work 2026, including the 503 South African respondents, job-displacement concern and job satisfaction against global averages). https://www.zawya.com/en/economy/africa/south-africas-ai-boom-is-already-changing-jobs-w3szunvo
- McKinsey & Company: Superagency in the Workplace: Empowering People to Unlock AI's Full Potential (28 January 2025; 3,613 employees and 238 C-level executives surveyed October to November 2024, US findings). https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work
- Harvard Business Review: Why Employees Aren't Transparent About Their AI Usage (June 2026). https://hbr.org/2026/06/why-employees-arent-transparent-about-their-ai-usage
- Microsoft and LinkedIn: 2024 Work Trend Index Annual Report: AI at Work Is Here. Now Comes the Hard Part (May 2024; 31,000 respondents across 31 countries). https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part
- CNBC, reporting the same study (8 May 2024), for the specific finding that 53% of those using AI on their most important tasks worried it made them look replaceable. https://www.cnbc.com/2024/05/08/workers-hiding-ai-use-on-important-tasks-over-fears-about-being-replaced-report-.html
Written by JP, Sixees Labs. Last reviewed August 2026.