IBM Study Finds AI Is Eroding Workforce Critical Thinking

There is a growing disconnect between what business leaders want from their AI-enabled workforce and what employees are actually developing. IBM’s Institute for Business Value just published its most comprehensive AI workforce study to date, and the findings should concern every business owner deploying AI right now.

The study surveyed 1,500 Chief Human Resource Officers across 21 countries and 8,800 full-time employees across 28 countries. The methodology, conducted in partnership with Oxford Economics from April to June 2026, makes this one of the most credible data points on AI and workforce skills we have right now.

The Gap That Could Sink Your AI Investment

Here is the number that should stop you mid-sentence: 71% of CHROs say the ability to supervise, validate, and override AI outputs is the most essential skill for their workforce. Yet only 29% of employees rank human judgment as important.

That is not a small gap. That is a fundamental misalignment about what it means to work with AI.

HR leaders are thinking about AI governance. They want people who can catch when an AI gets it wrong, who can interrogate AI outputs before acting on them, who know when to say “no, that does not seem right.” Employees, meanwhile, are focused on using AI faster and more efficiently, treating it more like a search engine than an assistant that needs oversight.

Neither group is wrong, exactly. But the combination is a recipe for expensive AI failures.

60 Percent of Employees Worry AI Is Making Them Worse

The skills erosion numbers are worth sitting with. Sixty percent of employees say they worry that AI is eroding their skills, and critical thinking is the one cited most often as declining. Among employees who are concerned about skills erosion, three in four say AI has already begun to erode at least some of their capabilities.

That is not hypothetical. People are feeling it in their daily work.

There is an important distinction here between using AI to do a task and using AI to do a task well. If a data analyst lets an AI tool build their dashboards without engaging with the underlying data, they are getting faster outputs but losing the ability to spot when the data is wrong. If a marketer uses an AI to write all their briefs without challenging its reasoning, they are producing faster content but losing their editorial instinct.

The IBM study found that CHROs identify critical thinking (57%) and human judgment (48%) as the capabilities they need most from their teams. Those are exactly the skills that atrophy fastest when AI is used as a shortcut rather than a collaborator.

When You Build Judgment Into the Work, It Shows

The research also points to what happens when organisations take this seriously. Where judgment is deliberately built into AI-enabled workflows, 62% of CHROs report growing employee confidence in AI-assisted decisions. Where it is not, 57% report confidence declining.

This matters because confidence in AI decisions is not just a soft metric. It is the difference between a team that catches an AI model’s error before it goes to a client and a team that passes it through because “the AI said so.”

The question is not whether to use AI. The question is how to structure work so that human oversight remains genuine rather than performative.

What This Means for Business

If you are a business owner or team leader deploying AI tools, this study raises three practical questions worth answering honestly:

Does your team know when to override the AI? Not just theoretically, but do they have the training and the permission to do it? Some organisations create cultures where questioning AI output feels like slowing things down. That is dangerous.

Are you measuring skill development, not just output speed? If your AI rollout is making teams faster but not developing their ability to evaluate, question, and improve AI output, you are building a dependency rather than a capability.

Is your AI deployment designed around human judgment or designed around removing it? There is a real difference between AI that augments human decision-making and AI that replaces it without adequate oversight structures.

The IBM research is a useful reminder that the hardest part of AI transformation is not the technology. It is the organisational design, the training, and the cultural shift required to make human-AI collaboration actually work.

Skills erosion ranked as a top concern for 46% of CHROs in this study. That is nearly half of HR leaders at major organisations, all worried that the tools meant to help their teams are quietly making those teams less capable.

The businesses that get this right will have AI-enabled teams who can catch errors, improve outputs, and exercise genuine judgment. The businesses that get it wrong will have fast, confident, and occasionally disastrously wrong AI outputs with no human circuit-breaker in place.

The Training Gap Is a Real Opportunity

For data and AI professionals, this research points to a skill set that is genuinely in demand: the ability to evaluate AI outputs critically, to understand the data behind a model’s recommendations, and to know when the AI is right and when it is not.

That is not just a technical skill. It requires data literacy, analytical thinking, and a working knowledge of how AI systems make decisions. It is also something that can be learned, practiced, and improved, which is exactly where structured training programs make a difference.

The enterprises winning with AI right now are not the ones who gave everyone a Copilot subscription and called it transformation. They are the ones who invested in building genuine human-AI collaboration skills across their teams.

---

This analysis is based on the IBM Institute for Business Value CHRO study released September 21, 2026, conducted in partnership with Oxford Economics. The full report is available at IBM’s newsroom.