Core takeaway: The productivity gains from AI are real, but they are being delivered in a way that systematically strips workers of the very things confidence is built on—training, agency, and a sense that they still own their craft.
The Paradox in One Number
ManpowerGroup’s 2026 Global Talent Barometer, based on interviews with nearly 14,000 workers across 19 countries, found that regular AI usage jumped 13 percentage points to 45% of the workforce—while confidence in using technology fell by 18%. It was the first overall decline in worker confidence in three years. The pattern is not confined to one sector: a 2026 survey of tech workers found that 82% say AI makes them more productive, yet burnout climbed from 44.7% to 55.7% in a single year. Productivity and confidence are moving in opposite directions. The explanation is not that AI fails to deliver output. It is that the way it is being deployed erodes the psychological foundations of competence.
The Training Void
The most immediate driver is structural neglect. More than half of the global workforce—56%—reported receiving no recent skills development, and 57% had no access to mentorship, even as AI use accelerated. “Workers are being handed tools without training, context, or support,” ManpowerGroup’s VP of global insights told Fortune. Confidence does not come from using a tool; it comes from understanding it. When organizations deploy AI as a mandate rather than as a capability-building investment, they produce adoption without mastery—and mastery is what confidence tracks.
Cognitive Debt and the Fear of Losing the Plot
There is a deeper mechanism at work. A randomized controlled trial with 1,222 participants found that AI assistance improves short-term performance but significantly impairs unassisted performance—people perform worse without AI and are more likely to give up. Researchers describe this as “cognitive debt”: the MIT study Your Brain on ChatGPT documented reduced neural activity during AI-assisted writing, and follow-up analysis confirms that passive delegation carries genuine costs, including shallower argumentation and skill decay. A tech worker quoted in one 2026 survey put it bluntly: “My brain is rotting. I just follow Claude. I don’t understand what I merge”. Confidence presupposes a self that is still capable without the tool. When the tool quietly replaces the self, the productivity gain arrives with an existential bill.
Algorithm Aversion and the Loss of Volition
Behavioral research adds another layer. People exhibit “algorithm aversion”—a reluctance to rely on algorithmic output even when it outperforms human judgment—and the effect intensifies when following the algorithm feels like a concession of personal agency. AI systems are optimized for instant, complete answers; they do not scaffold the experience of working through a problem, and that scaffolding is precisely what builds durable skill and the confidence that accompanies it. Workers are not irrational to feel less confident. They are accurately registering that the locus of competence has shifted away from them.
Surveillance Turns Productivity Into Suspicion
Finally, the measurement apparatus around AI corrodes trust. A meta-analysis of electronic performance monitoring found no evidence that it improved worker performance—and instead linked it to increased stress. When monitoring is invasive or opaque, “trust erodes,” and employees under surveillance are 1.5 times as likely to report poor mental health. Productivity metrics that feel like surveillance transform a tool for doing work into a tool for being judged, and confidence cannot flourish under constant evaluation.
What Would Rebuild Confidence
The decline is not a verdict on AI’s value. It is a verdict on deployment. Confidence rises when workers are trained, when AI is framed as augmentation rather than replacement, and when monitoring is transparent and limited. The productivity is real; the confidence deficit is a design failure, not a technological one.
