60% of employees fear AI is eroding their skills, IBM finds
IBM’s new CHRO study says AI skills erosion is now a workforce concern, with critical thinking and human judgment emerging as priority capabilities for the next phase of work.
AI is making work faster. But are workers becoming less skilled along the way?
That question is becoming harder to ignore as AI moves deeper into everyday jobs. A new global study by the IBM Institute for Business Value found that 60% of employees fear AI is eroding their skills, with critical thinking emerging as the biggest concern.
The study, released on September 21, 2026, surveyed 1,500 Chief Human Resource Officers and equivalent senior executives, along with 8,800 full-time employees across global markets.
The findings reveal an interesting gap. Companies want people to make better decisions with AI, but many employees are more concerned about whether they are retaining the ability to make those decisions themselves.
The skill gap AI is creating
For CHROs, judgment is becoming one of the most important workplace skills. IBM found that 71% of CHROs consider the ability to validate, supervise, and override AI outputs essential. Only 29% of employees, however, ranked judgment as important.
Critical thinking and problem framing were also high on the list for HR leaders, with 57% identifying them as important skills for the AI era. Human judgment followed at 48%.
Now for employees, the concern is already showing up in their day-to-day work. Among those worried about skills erosion, three in four said AI had already begun weakening at least some of their skills.
The issue is not necessarily that AI is replacing every task. It is that when AI handles more of the thinking, checking and execution, workers may get fewer opportunities to practise those abilities themselves.
The work AI creates behind the scenes
AI can also make work more complicated even when it appears to make it faster. IBM found that 80% of CHROs believe AI creates “invisible” work for employees. That can include checking AI recommendations, correcting errors, adding missing context and handling exceptions.
Workers are noticing the extra load. About 42% said AI either increases their workload or creates additional work that goes unrecognised. Another 43% said they are blamed when something goes wrong with AI.
That creates a different kind of productivity problem. An employee may finish a task faster with AI, but still spend significant time checking whether the output is accurate and deciding what should happen next.
Who decides when AI gets it wrong?
This is where workplace design becomes important.
Companies need to be clear about which decisions remain human-led, which are AI-assisted and which can be executed by AI. Without that clarity, employees can end up responsible for outcomes without having clear control over the systems producing them.
IBM found that organisations with clearly defined human-led, AI-assisted and AI-executed workflows reported 18% lower risk and 20% higher quality. Yet HR is not always part of the conversation. The study found that 46% of organisations do not involve their CHRO when defining AI strategy.
HR teams are also still early in adopting AI themselves. About 72% of organisations have limited or no use of AI within HR, while CHROs rated their organisations relatively low on AI literacy, AI performance measurement and change management.
The new workplace skill may be knowing when not to trust AI
The findings point to a different way of thinking about AI training. Learning how to use an AI tool is only one part of becoming AI-ready. Workers also need to understand when an output should be questioned, how to identify errors and where human judgment needs to remain in control.
As AI takes over more routine work, critical thinking, problem framing and accountability may become more important precisely because there is less practice built into the work itself.
The challenge is not simply faster AI adoption. It is ensuring people retain the skills to understand, challenge and improve AI-generated work.