AI Is Doing the Work but Are Your Leaders Still Doing the Thinking

Fast Company recently asked a question that lands hard for anyone who manages people in 2026: if AI is doing the work, are your leaders still doing the thinking? The short answer is that many are not, and the data explains why. As AI systems begin to initiate tasks, execute them, and report back, the human role can quietly slide from creator to approver. At Metaintro, we watch this play out across the hiring market every day, and the pattern is clear: the professionals who stay valuable are the ones who keep owning the thinking even when the machine offers to do it for them.

What Changed When AI Started Doing the Work Itself?

For most of the last decade, workplace software waited for a human to press go. Fast Company argues that AI has now crossed a threshold where it no longer simply helps people do their jobs but acts as something closer to an agent that starts a task, carries it out, and hands back a finished result. That sounds like pure productivity, and in narrow ways it is. The problem is that the decision loop, the cycle of framing a question, weighing options, and choosing a path, used to belong to a person. When AI moves through that loop on its own, the leader can end up reviewing an output rather than shaping the reasoning that produced it.

This is a bigger change than a faster spreadsheet. It touches the core of what management is for. A leader who only signs off on machine-generated plans is no longer setting direction, and over time that erosion is hard to notice because the work still ships and the dashboards still look green. We see a related dynamic in how companies adopt these tools, which we covered in our look at why so many US companies remain stuck in an experimental AI phase. The technology arrives fast, but the thinking discipline around it lags behind.

Why Does the Confidence Paradox Put Leaders at Risk?

The most useful evidence here comes from a study by Microsoft Research and Carnegie Mellon University called The Impact of Generative AI on Critical Thinking. The researchers surveyed 319 knowledge workers who described 936 real workplace uses of generative AI, giving them nearly 1,000 concrete examples to analyze. The headline finding is what researchers describe as a confidence paradox: the more a person trusted the AI tool, the less critical thinking they applied to its output, while the more a person trusted their own ability, the more they scrutinized what the machine produced.

That finding should worry anyone who leads a team. It means the natural human response to a smooth, confident answer is to relax, not to probe, and confident answers are exactly what modern AI is built to produce. Across the cognitive tasks the study tracked, workers reported spending much less or less effort in 72 percent of knowledge tasks, 79 percent of comprehension tasks, 76 percent of synthesis tasks, and 55 percent of evaluation tasks. The reporting of the study, covered by Forbes, notes that critical thinking itself is shifting from information gathering to information verification and from task execution to task stewardship. For leaders, that shift is the whole job now. If nobody is verifying and nobody is stewarding, the organization is running on autopilot with a very persuasive co-pilot.

Which Human Skills Get More Valuable as AI Does the Execution?

If AI absorbs the execution, value moves to the skills that sit upstream and downstream of it. The first is framing, the ability to define the real problem before any prompt is written. A machine will happily answer the wrong question in beautiful prose, so the person who can state the objective clearly is worth more, not less. The second is judgment under ambiguity, the willingness to weigh trade-offs the model cannot see because they involve people, timing, and context that never made it into the training data. We explored a version of this in our piece on why five skills now beat your job title in AI workplaces, and the same logic applies at the leadership level.

There is a third skill that quietly separates strong leaders from passive ones, and that is the willingness to disagree with a fluent answer. A model is engineered to sound sure of itself, and certainty is persuasive even when it is wrong. The leader who can say the reasoning looks thin, ask for the assumptions behind a recommendation, and send the work back for another pass is doing something the machine cannot do for itself. That skill is not about technical fluency at all. It is about intellectual courage, and it becomes rarer and more valuable precisely as the outputs around it get slicker. A related human strength is taste, the earned sense of what good looks like in a specific domain, which lets a person spot when a plausible answer is subtly off. Taste comes from experience and attention, and it is one of the few advantages that compounds rather than depreciates as tools improve.

The Fast Company argument leans on the psychologist Albert Bandura, who described four properties that make people genuine agents of their own behavior: intentionality, forethought, self-reactiveness, and self-reflectiveness. Unchecked agentic AI erodes all four, because it removes the friction where those qualities normally get exercised. That makes intentionality a leadership skill worth naming out loud. So is curiosity, which we argued matters more than a fancy credential in our story on the leadership skill that beats an MBA. The durable skills are stubbornly human, a point reinforced by our reporting on the jobs AI still cannot replace and on how adaptability now beats raw experience.

How Does a Leader Keep Doing the Thinking Day to Day?

The practical fix from the Fast Company piece is to reintroduce intentionality before touching the tool. Before engaging AI on any task that matters, a leader should articulate their objective, state their own perspective, and decide how the AI will support their thinking rather than replace it. That sounds simple, but it inverts the default habit of opening a chat window and asking the machine what to do. Stating your view first means the AI output becomes something to test against, not something to accept.

A second habit is to keep a visible human step in every important decision loop, a moment where a person must explain the reasoning, not just approve the result. This is where the confidence paradox gets neutralized, because the act of explaining forces the scrutiny that trust would otherwise suppress. Leaders can also rotate who plays the skeptic, so disagreement becomes a role rather than a personality trait. These are small rituals, but they protect the thinking. We have seen the alternative in organizations where nobody could connect the tooling to the work, a gap we documented in our reporting on the wide disconnect between AI training and workers' actual jobs.

What Does This Mean for Your Career if You Are Not a Leader Yet?

Here is the career angle, and it matters whether you manage anyone or not. The market is repricing skills in real time, and the people who become passive approvers of AI output are making themselves easy to replace, because approving is exactly what the next tool will do more cheaply. The people who get more valuable are the ones who can frame problems, challenge a confident wrong answer, and own a decision end to end. If you want to future-proof your path, invest in the judgment layer, not just the tooling layer, an argument we made in our guide to career skills that keep you employable through constant change.

Practically, that means treating every AI-assisted task as a chance to practice verification rather than a shortcut to skip it. Read the output like an editor, not a rubber stamp. Ask where the reasoning could be wrong and whether the framing was even right. This is also why emotional intelligence and coaching skills are rising in value, because they cannot be automated and they compound over a career. The workers who thrive in an AI-heavy workplace are not the fastest prompters. They are the clearest thinkers, and clear thinking is a habit you can build on purpose.

Are Managers Becoming Obsolete or More Important?

It is tempting to conclude that if AI can plan, draft, and decide, then the manager is a redundant layer. The evidence points the other way. When execution gets cheap and abundant, the scarce resource becomes good judgment about which work is worth doing at all. That is a management function, and it gets harder, not easier, when a machine can generate ten plausible options in seconds. Someone has to choose, defend the choice, and own the outcome. We dug into this tension in our coverage of how accidental managers sink their teams, and the AI era raises the stakes on every point in that story.

The risk is not that managers disappear but that they hollow out, keeping the title while outsourcing the thinking. A leader who lets AI set priorities is not saving time, they are quietly handing over the part of the job that justified the role. Even executives are wrestling with where the line sits, as the Costco chief executive signaled when he said AI is not making the decisions in his retail business. The manager who stays valuable is the one who uses AI to widen the option set and then does the human work of choosing well.

How Should Teams Redesign Work So AI Supports Thinking Instead of Replacing It?

The organizational answer is to design workflows that keep humans in the reasoning loop by default rather than as an afterthought. That means building in checkpoints where a person states the goal before the AI runs, and review steps where someone must articulate why an output is right before it ships. It also means measuring the quality of decisions, not just the speed of output, because speed is exactly what makes the confidence paradox dangerous. When a team rewards fast approvals, it trains itself to stop thinking. When it rewards good reasoning, it keeps the muscle strong.

There is also a cultural layer that leaders control. If a team treats every AI answer as a starting draft rather than a verdict, the thinking survives. If it treats the machine as an authority, the thinking quietly leaves the building. Leaders set that norm through what they praise and what they question in public. When a leader visibly pushes back on a slick output, asks where the reasoning came from, and rewards the person who caught a flaw, the whole team learns that verification is the real work. When a leader waves everything through because it looks polished, the team learns the opposite lesson fast. Culture, in other words, is either a defense against the confidence paradox or an accelerant for it, and which one it becomes is a leadership choice made a hundred small times a week.

Hiring plays a role too. As execution becomes commoditized, employers are chasing the specific human skills that AI cannot supply, a shift we tracked in our reporting on how employers now chase specific skills over generic experience. The teams that win will treat AI as a thinking partner that must be interrogated, not an oracle to be obeyed. That is the whole lesson of the confidence paradox, and it is why the leaders who keep doing the thinking will outlast the ones who quietly stopped. The work is changing fast, but the value of a clear human mind is going up, not down, a theme that runs through our analysis of AI's biggest impact landing on minds rather than jobs.

Lacey Kaelani, CEO of Metaintro, told Authority Magazine, "Stop trying to predict the future and start building adaptability into your systems." That is the antidote to the confidence paradox, because a leader who builds in checkpoints for human judgment, rather than betting on any single AI answer, keeps the thinking alive as the tools get more persuasive.

Related Articles

• AI's Biggest Impact Is on Minds Not Jobs
• The 2026 Leadership Skill That Matters More Than an MBA
• Five Skills That Beat Your Job Title in AI Workplaces
• The 2026 AI Acumen Gap and the Boss Test
• Career Skills to Stay Employable Through Constant Change
• Seven Jobs AI Cannot Replace and What Each Pays
• Adaptability Beats Experience for Career Growth
• How to Coach a Team on Emotional Intelligence
• The 2026 AI Paradox of Replacing the Experts It Needs
• Accidental Managers Sink Teams
• Harvard Study on AI Reshaping Jobs as Workers Adapt
• Agentic AI Career Skills Define the Next Wave of Hiring

People Also Asked

Q: Does using AI actually reduce critical thinking?

A: Research from Microsoft and Carnegie Mellon University found a confidence paradox, meaning the more workers trusted the AI tool, the less critical thinking they applied to its output. Across the tasks studied, most workers reported spending less effort, which suggests the risk is real when people treat AI answers as final rather than as drafts to be checked.

Q: What leadership skills matter most as AI does more of the work?

A: The skills that gain value are the human ones AI cannot originate, including framing the right problem, exercising judgment under ambiguity, and holding intentionality about the goal before any tool runs. The Fast Company piece frames this around stating your objective and perspective first, so the AI supports your thinking rather than replacing it.

Q: Will AI make managers obsolete?

A: The evidence suggests the opposite. When execution becomes cheap and abundant, good judgment about which work is worth doing becomes the scarce resource, and that is a management function. The danger is not managers disappearing but hollowing out by keeping the title while outsourcing the thinking to the machine.

Ready to level up in an AI-driven workplace? At Metaintro, we help you find roles that reward judgment and human skills, not just speed, so you can build a career that AI makes stronger rather than smaller. Explore verified opportunities and sign up here to get matched with employers who value how you think.