AI Is Taking Over Workplace Critical Thinking

Most leaders believe their teams are using AI as a tool. The research suggests something more consequential is happening. Workers are not just using AI to work faster. They are letting it decide, and in doing so, quietly ceding the human reasoning that determines whether those decisions are any good.

A January 2026 paper from the Wharton School introduced a term for what is now documented and measurable. Researchers Steven Shaw and Gideon Nave call it cognitive surrender.

They define it as adopting AI outputs with minimal scrutiny, thereby overriding both intuition and deliberation. Their framework extends Nobel Prize-winning psychologist Daniel Kahneman's model of fast intuitive thinking and slow deliberate thinking by introducing a third system: artificial cognition that operates entirely outside the brain.

That third system, they argue, can supplement or supplant human reasoning. When it supplants it, AI stops being a thinking partner and starts being the decision-maker.

The performance implications are direct. When workers in the study consulted an AI that was correct, their accuracy rose significantly above what they achieved on their own. When the AI was wrong, their accuracy fell well below the baseline of people who had no AI access at all. The problem is that workers had no reliable way to detect the difference. They accepted incorrect AI answers 80 percent of the time. Their confidence rose either way, whether the AI had helped them or led them astray.

This dynamic is particularly consequential with the large language models now embedded in most workplace tools. LLMs do not retrieve facts. They generate plausible-sounding responses based on patterns in training data, without access to an organization's specific context, strategy, institutional knowledge, or the domain expertise of the person using them. They do not flag uncertainty. They speak with consistent confidence regardless of accuracy.

Getting strong output from an LLM requires a skilled human on the other end: one who validates what it produces, identifies what it missed, expands the ideation beyond the initial response, and applies judgment to decide. Cognitive surrender eliminates every one of those steps.

A Microsoft Research study published in April 2025 found that confidence in AI was among the strongest predictors of whether knowledge workers engaged in critical thinking at all. The higher the trust in the tool, the less scrutiny is applied to what it returned.

As researchers noted in that same study, there is a fundamental irony at the center of automation: when routine cognitive tasks are mechanized and handed to an external system, the human is deprived of the routine practice that builds and sustains judgment. The reps disappear. And so, over time, does the muscle.

A McKinsey "State of Organizations 2026" report, published in February, found that only 23 percent of organizations qualify as AI Pioneers, those actively deploying AI across most departments and functions with a clear understanding of how it will reshape their work. The vast majority are still experimenting, running isolated pilots, or deploying AI in piecemeal ways that have yet to generate measurable enterprise-wide impact.

That gap is the opportunity. Cognitive surrender is not yet the organizational norm. Supplementing human reasoning rather than supplanting it is still a choice leaders can architect into how AI gets deployed. The window to make that choice intentionally, before passive AI reliance becomes the default operating culture, is shorter than most leaders assume.

The structural interventions the research points to are specific. Leaders who build these practices into how AI gets deployed now safeguard their organizations from a culture where AI supplants human reasoning rather than supports it, and from the performance costs that follow.