A senior leader sits in a boardroom in early 2026. On the screen in front of him is a strategy brief his AI co-pilot helped him build in forty minutes. The presentation is clean. The numbers are crisp. The language sounds like him at his sharpest.
Three of the people sitting around him have noticed something the AI did not. They are not going to mention it.
This is the modern shape of an old leadership failure. I have been writing about it for years, in a different form. Let me tell you about a leader I worked with named William.
A Surgeon Walks Into a Department
William was a cardiac surgeon. After years in the operating room, he took a position at a new hospital as medical director of a department. The role meant splitting his time between the OR and overseeing the work on the floor. He had been considering the move for some time. The only thing he had not thought about was his leadership style.
It did not take long for William to notice something was off. His new team seemed almost eager to shoot down his ideas. Others avoided him. He had a constant sense that people were talking about him behind his back.
Then the system-wide survey came back. Beside his name was a single phrase: Most Hated Physician.
William was stunned. As a surgeon, he had built his identity on a particular mindset. In the operating room, the accountability was his. He had earned the right to run the show his way. My operating room. My anesthesiologist. My instruments. My way or the highway.
That mindset made him excellent in the OR. It made him hated on the floor.
The new role required collaboration. The very smart people who worked alongside him every day knew things he did not, and saw things he could not see from his position. Walking in with “my way or the highway” closed every door before anyone reached for the handle.
What William Had to Learn
What changed William, six months later, was a framework he had been studying in a leadership program. The Three Lenses, developed by colleagues at MIT Sloan, asks a leader to view the same situation through three different perspectives.
The Strategic lens: how does the work of this group actually serve the organization? What processes work? What does not?
The Political lens: who holds the knowledge here? Who holds the positional power? How are decisions actually made?
The Cultural lens: what stories live inside this team? What does the team value? Where is it safe to speak, and where is it not?
William stepped back. He observed. He asked. He learned to listen for what his team knew that he did not. He apologized to the people he had treated badly. He started having real conversations.
Six months later, the same survey returned a different result. William was now the Most Admired Physician in the system.
He had not learned a new skill. He had unlearned an old one.
How AI Is Recreating William's Trap
I am writing about William again because I am watching the same trap close around a new generation of leaders.
Microsoft’s 2025 Work Trend Index found that “78% of leaders trust AI for high-stakes work,” compared with 66% of employees. Read that gap carefully. It is the modern version of William in the OR. The leader is more confident in the tool than the people who have to live with the decisions the tool helped make.
A separate Microsoft Research study of 319 knowledge workers, published in early 2025, put a sharper edge on the problem. It found that “higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking.” The more a leader trusts the AI, the less the leader scrutinizes what the AI produced. The fewer questions get asked. The faster a confident wrong answer reaches the room.
There is a phrase I keep hearing when leaders defend their reliance on AI: but it’s right most of the time. That is exactly the danger. William’s surgical mastery was right most of the time too. The few times he was wrong in the OR, his system absorbed it. Outside the OR, on the floor where decisions were made by negotiation, his “rightness” became a wall.
Today, the equivalent of William’s wall is the leader who walks into a meeting with an AI-built brief and treats it as the answer. The team in the room has the same instinct William’s nurses and doctors did. They know the leader is no longer asking. So they stop telling.
The Three Lenses, Refreshed for the AI Era
The Three Lenses framework was built for a world in which a leader’s perspective was the bottleneck on understanding. AI has not removed that bottleneck. It has added a fourth voice in the room, often the loudest one. Each lens shifts when AI is functionally a member of the team.
The Strategic Lens. When AI can produce a confident answer in forty minutes, the strategic skill of leadership is no longer finding the answer. It is asking the right question, and listening for what the people in your team see that the model cannot. The competitive edge is not in the brief. It is in the conversation around the brief.
The Political Lens. When a leader cites an AI output as the basis for a decision, the people in the room who would have pushed back on a peer go quiet. Authority transfers to the machine. Most leaders never notice it happening. The moment the team stops debating with you, the political lens has already failed.
A BCG study from July 2025 found that 46% of employees worry about decisions being made without human oversight, and 35% are concerned about unclear accountability when mistakes occur. The political risk is not abstract. It is sitting in the chairs across from you.
The Cultural Lens. Researchers at Stanford’s Social Media Lab and BetterUp Labs published a study in late 2025 that named the cultural cost: workslop. Polished, AI-generated material that the recipient then has to clean up because the sender did not invest the thought. 40% of workers reported receiving it in the past month. They responded with annoyance, frustration, and a quietly damaging conclusion: that the colleague who sent it had not thought about them enough to do their own thinking.
Trust does not collapse in a single moment. It erodes one prompt at a time.
What I Am Inviting Senior Leaders To Do
When I think about how to coach leaders through the AI-confidence trap, the practice has three parts.
Hold the AI output as a draft, never as the answer. Ask the same questions you would ask of a junior analyst’s first cut. What is missing? What did the model not know? What is the assumption underneath the recommendation that someone in this room would challenge?
Bring the team into the conversation before you bring in the conclusion. The political lens is most easily lost when a leader has decided before the meeting begins. If you walked in with the AI’s answer in hand, you have already closed the room.
Notice the silence. When your team stops pushing back, that is not consensus. It is the signal William got too late. The Most Hated Physician was not the surgeon who was wrong. He was the surgeon who had stopped asking.
An Invitation
When I first wrote about William, the trap was a leadership-style problem that most leaders only inherited if they came from a deeply technical background like surgery. AI has democratized the trap. Any leader can fall into it now, in any meeting, with any prompt.
The technology around us has changed enormously since the day William got his Most Hated label. The leadership pattern has not changed at all. The new “my way or the highway” sounds like polished AI output and looks like decisive leadership. It produces the same survey result.
I invite you to ask yourself: when did you last walk into a meeting having decided, and how would your team know if you had?
On with the show.
Rob Salafia is an executive leadership coach, lecturer at MIT Sloan School of Management, and author of Leading from Your Best Self (McGraw-Hill). He is the founder of Protagonist Consulting Group and has spent more than 25 years developing executive presence in leaders at companies including AstraZeneca, American Express, ING Bank, Sony Music, and Johnson & Johnson.
