As executives plan AI-driven headcount cuts, some employees are pushing back against the rollouts meant to replace them. In a survey of 2,400 knowledge workers, 29% admitted to sabotaging their company's AI strategy, with nearly a third of those citing fear for their jobs. The resistance reflects a broader sentiment that AI is being imposed on employees and not introduced for their benefit. Companies seeing returns from AI recognize that employee involvement is essential, giving people a meaningful role in how the technology is introduced and used.
Human barriers and how to overcome them
Every transformation we run encounters two human barriers:
Imagination: Leaders and teams struggle to picture how work could be done differently.
Digestion: Once the new system is in place, people have to change how they work, and change is uncomfortable.
We saw the second barrier clearly at a customer support organization where the new system could handle roughly 95% of the team's volume. The team resisted handing it over and held transactions back for themselves, believing that doing the work manually protected their value.
Experience taught us that the people who step back from doing the work and take responsibility for managing the system become the most valuable. Leaders can support that shift by defining people's roles before the system goes live.
When IKEA introduced its Billie AI assistant to handle routine customer inquiries, the company retrained roughly 8,500 call center employees to manage more complex service issues or become interior design sales advisors. Its remote-sales channel has since become its fastest-growing, generating €1.25 billion in annual sales.
A mid-market recycling company we work with took a similar approach. After we reimagined its operations workflows and increased capacity more than fivefold, the company chose not to lay off anyone. Attrition will account for some workforce changes, while some operations staff are moving into sales, where personal relationships drive revenue.
In both cases, employees were given new responsibilities as AI took over existing tasks. Defining those roles early gives people a clearer picture of their future with the company.
The double standard of human and machine performance
At Techjays, we often launch new AI systems for clients that achieve 96% accuracy on day one. Leadership often comes back and says it's not good enough. Then we ask about the human team's accuracy rate. It's 91%. The machine is already five points better and still learning, but for some reason people hold machines to a standard of perfection they never apply to each other.
Researchers at Wharton call this phenomenon algorithm aversion. People lose confidence in an algorithm faster than in a human after watching both make the same mistake, and they often choose the human even after seeing the algorithm outperform. Self-driving cars show the same pattern at scale. Waymo, for example, reports 82% fewer injury-causing crashes than human drivers across more than 270 million autonomous miles, and an independent IIHS analysis found Waymo vehicles involved in 68% fewer crashes per mile than people. Yet every robotaxi incident makes headlines.
In a follow-up to the Wharton study, the same researchers found people were far more willing to use an imperfect algorithm when they could adjust its output, even slightly. We build that finding into our approach deliberately. When someone corrects a drafted email or an estimate, the correction feeds the evaluation system, the agents learn from it, and the people doing the correcting can see their judgment shaping the system.
AI redefines the value humans bring
At a health benefits organization we work with, people who once processed claims one by one now oversee roughly ten times as many, stepping in only on cases that need human judgment. Those cases include unusual claims, angry customers, bids with unfamiliar specifications, and decisions that depend on personal relationships.
This new approach changes what makes a great employee. Previously, performance was measured by how many tickets someone processed or how many lines of code they wrote. Volume was the priority. Now, judgment, taste, and the ability to spot subtle errors in AI-generated work matter more. Domain expertise becomes even more valuable because it allows people to recognize when an output looks right but isn't.
Companies should rethink how they hire, train, and evaluate employees accordingly:
Hire for judgment and curiosity over throughput. Speed is now the system's job.
Grow the skill internally. People who can direct agents and own the exceptions barely exist in the labor market yet. Your best current operators are the most likely candidates.
Rewrite roles and incentives. Most job descriptions, performance metrics, and compensation plans were designed for people doing the work, not overseeing it. Leaving them unchanged rewards the old behavior.
Managing resistance to transformation
A PE-backed software company with a 12-person engineering team came to us recently. Six of the engineers flatly refused to adopt agentic engineering. They'd been with the company for years, and the CEO wanted to give everyone a fair chance.
This is more common than most leaders admit. In Stack Overflow's 2025 developer survey, 84% of developers used AI tools, yet 46% distrusted their accuracy. Skepticism isn't the same as obstruction, and the response should account for that.
Establish a baseline instead of debating. We evaluate the codebase and measure current delivery, so the conversation is about numbers rather than opinions.
Build with the willing. The engineers ready to work in an AI-native way take on a scoped rewrite or a real backlog item, on the edge of the organization.
Make results visible. Side-by-side output, cycle time, and quality speak louder than any mandate.
Turn resistance into a role. The engineer most suspicious of AI-generated code is often the best reviewer of it. Give skeptics the verification seat.
Be clear about where this is going. A well-run AI-native team of 12 can deliver what once took far more people, and the job is shifting from writing code to directing agents and applying judgment. Everyone deserves to know that, along with a timeline.
Keeping humans in the loop allows companies to make better use of the judgment, expertise, and relationships that AI can't replicate. Companies that account for these strengths from the start can improve performance while giving employees a meaningful role in the new way of working.
