AI Governance in the Enterprise | Article 4
Why is the implementation of AI primarily a leadership task? This article highlights four human friction points in AI adoption—and explains why true enablement goes far beyond tool training and prompt workshops.
In most companies, introducing AI is treated like a rollout: select the tools, train employees, communicate the benefits—and assume the rest will take care of itself. Experience tells a different story. The technology is rarely the problem. The problem is the people who are expected to work with it. And people cannot simply be rolled out.
Why is the introduction of AI a leadership task?
The introduction of AI transforms not only processes and tools but also roles, performance dynamics, collaboration, and professional identity. Therefore, simply rolling out technologies and training employees is not enough. Leadership must shape how people learn to work with AI, how work is redistributed, how uncertainty is managed, and which skills the company will need in the future.
In the previous article, I described how the leadership role changes when AI takes over operational administration—and how one of the tasks that remains is leading people through change. That is exactly what this article is about. Because this is where it becomes clear whether leadership in the age of AI has substance or was merely a flurry of activity.
The common assumption is that implementing AI is essentially a communication task. Explain it well enough, take people’s fears seriously, offer training—and they will get on board. On the surface, this assumption is correct. But it misses what is happening underneath.
Beneath the surface, fault lines emerge that cannot be communicated away. I consider four of them particularly consequential. Each one determines whether a leader shapes what happens—or merely manages it.
What problems arise during the implementation of AI?
In practice, four particular points of divergence emerge:
• Performance disparities: Employees benefit from AI to varying degrees, and gaps in competence can widen.
• Change in the nature of work: Work that involves shaping or creating can increasingly shift towards work that involves monitoring and correcting.
• Loss of the culture of questioning: People rely more heavily on AI answers and may question results less.
Real fears of loss: roles, tasks and sometimes even jobs can change or disappear.
Fault line one: The workforce splits—and predictably so
The first fault line is often overlooked because it looks like success. AI makes employees more productive—but not all of them, and not to the same extent. Some team members confidently master the new tools and pull ahead. Others cannot keep up, fall behind or refuse to engage. Suddenly, a gap in performance and status runs through the workforce that did not exist in this form before.
This is more than a perception. A widely discussed study by researchers from Harvard and Boston Consulting Group—a carefully designed field experiment involving 758 consultants—showed just how unevenly AI affects performance. For tasks aligned with AI’s strengths, performance increased by up to forty percent. For tasks beyond that boundary, performance deteriorated—by around nineteen percentage points. The researchers call this technology’s “jagged frontier”: AI is not equally good at everything, but uneven—and consequently affects the people working with it in very different ways.
An independent study by the University of Konstanz confirms that this divide is real: AI use in Germany varies by occupational group and educational level—the gap is not closing; it is becoming entrenched. Another figure from the same survey should make leaders sit up and take notice: more than ninety percent of respondents report a gap between the officially announced AI strategy and what actually happens in day-to-day work. The divide therefore exists not only among employees, but also between leadership’s aspirations and reality.
At this point, it is worth considering who collects such figures. One frequently cited study by an AI provider claims that an overwhelming proportion of executives are deliberately building a class of “AI elite” employees, while many plan to dismiss those who do not use AI. Such figures must be treated with caution—the provider sells an AI platform and has an interest in promoting precisely this narrative. Yet that is also what makes the finding revealing: when even a source with a vested interest acknowledges that organizations are deliberately creating a two-tier workforce, and independent research points in the same direction, the phenomenon is difficult to dismiss.
And it has an uncomfortable downside. When employees use AI to produce polished but hollow work—reports that look impressive but fall apart on closer inspection—the actual work does not disappear. It is simply passed on to the more diligent colleagues, who must clean up the mess. A study by BetterUp and the Stanford Social Media Lab quantified this: around four in ten employees had received this kind of incomplete work from colleagues in the previous month, requiring almost two hours of correction per case. The divide is therefore not merely a status gap. It is a quiet conflict over workload distribution: some save time while others clean up after them.
This is where the real leadership question lies. In some organizations, this divide is intentional. In most, it simply emerges when AI is introduced in the name of productivity without actively managing the organization’s learning curve. Those who fall behind eventually stop daring to ask questions. Those who are ahead do not automatically share their knowledge. And those correcting other people’s incomplete work have less time to learn themselves.
What remains is a company in which AI may be used everywhere—but only superficially by many, because the knowledge of advanced users does not spread across the organization. This is not only a matter of fairness; it is untapped potential—and unlocking it is a leadership responsibility.
Fault line two: Work loses its core
The second fault line concerns something that is harder to measure than productivity but has a deeper impact: the meaning of work itself.
AI often does not relieve people of the tedious parts of their work, but of the demanding ones. The machine writes the text; the person reviews it. The machine produces the analysis; the person checks it. The person who once created something becomes the person who supervises it. And that supervision is work—invisible, thankless work.
A survey by the AI provider Glean produced a telling figure—again, the caveat regarding a vendor source applies, but the finding aligns with what I hear in conversations. On average, employees save around eleven hours per week through AI. They then spend approximately six and a half of those hours supervising, correcting and feeding context to the AI. In other words, well over half of the time saved is immediately consumed by work that AI does not perform but causes. The phenomenon already has a name: botsitting—sitting beside the bot and keeping an eye on it, much like a child.
The problem is not only the time involved. The problem is what this shift does to professional identity. A software developer who was passionate about elegant solutions fears becoming little more than a button-pusher. Professionals whose pride was rooted in their craft watch that craft migrate out of their role. The insidious aspect is that this erosion is not dramatic but gradual—until small shifts have created a situation in which someone no longer recognizes their own work. Research describes a paradox here: on the one hand, AI strengthens the feeling of accomplishing more; on the other, it increases uncertainty about one’s own value—both at the same time, in the same person. This explains why enthusiasm and withdrawal can be seen in the same team, often among the same people.
For leadership, this means that meaning is not a soft, feel-good topic reserved for good times. It is a hard factor in employee retention. The same Glean survey shows—as a correlation, not a proven causal relationship—that employees with a particularly heavy supervisory burden are significantly more likely to be looking for another job. Leaders who introduce AI solely for efficiency and overlook what it does to people’s experience of work risk pushing their best employees into a hollowed-out supervisory role – and losing them as a result.
Fault line three: The team stops asking questions
The third fault line is the quietest—and perhaps the most dangerous, because it affects the foundation of a learning organization: its culture of asking questions.
When a machine instantly provides a fluent, convincing answer to every question, it changes the way people deal with not knowing. Two effects interact here.
The first concerns the culture of scrutiny. An experiment by Harvard Business School involving more than two thousand participants, who assumed the role of loan officers, revealed something significant: eighty percent wanted to see the AI’s recommendation—but only just under half also wanted to see the reasoning behind it. People want the result, not the why. They avoid further inquiry as soon as it becomes uncomfortable. The study’s author sums it up in a way that extends far beyond lending: the greatest risk of AI is not that it gives wrong answers—but that it trains people to stop asking why.
The second effect concerns psychological safety. Harvard researcher Amy Edmondson has shown that teams learn only when their members feel safe admitting that they do not know something or have made a mistake. Learning, she says, is uncomfortable—it means acknowledging that you do not know something. That becomes harder when there is a machine in the room that seemingly knows everything. Anyone who now says “I don’t know” feels exposed twice over: after all, they could have asked the machine. Surveys show the consequence: a significant proportion of employees are now too embarrassed to ask for help with new technology—more than half among younger employees. AI intensifies this further because it never says, “I don’t know.” It always provides an answer, confidently worded, even when it is wrong. It models a false certainty against which people unconsciously measure themselves.
Linked to this is a phenomenon I write about with deliberate caution, as the evidence remains inconclusive. We are seeing younger employees increasingly turn to AI rather than experienced colleagues—describing their chats with the machine as conversations with “someone who knows everything” and reaching out to human peers less often. This potentially eliminates the very mechanism by which demanding professions have historically perpetuated themselves: learning through inquiry, correction, and guided practice. Yet, in fairness, we must also consider the other side of the coin. In the past, the seasoned colleague’s valuable advice primarily benefited those who were well-connected or bold enough to ask questions in the hallway; the shy or the new often missed out. AI, however, does not ask whether you “belong.” Perhaps, for the first time, everyone is gaining access to the kind of guidance that was once reserved for a select few.
It is an open bet—and that is precisely where leadership bears responsibility. Today, nobody knows whether organizations are eliminating an inefficient tradition or losing the quiet substance of their talent pipeline. The answer will only become apparent in five or ten years, when today’s experienced employees retire—and it becomes clear whether anyone has inherited their judgment. Making a conscious decision under this uncertainty, instead of simply letting events take their course, is not a technical task but a profoundly entrepreneurial one.
Fault line four: Some of the fear is justified
The fourth fault line is the most delicate on a human level. The usual response to employees’ fear of AI is reassurance: “Don’t worry, AI isn’t replacing anyone.” The problem is that sometimes this is not true—and people know it.
Anyone who sees a real loss approaching and is told that it is not such a big deal does not feel reassured, but dismissed. And a feeling that has been talked away does not disappear. It returns—as resistance, internal withdrawal or cynicism in the break room. Not because employees are irrational, but because they have been denied the right to experience a genuine loss as a loss. Psychologists refer to this as anticipatory grief: people mourn a future they believed was within reach before anything has actually happened. If this feeling is not given legitimate space, it finds another outlet.
Honest leadership means naming the loss instead of talking it away. No false comfort, but no alarmism either—simply acknowledging that something is changing and that this is painful for some people. Only this honesty allows a leader to become a credible point of contact again for what comes next. It also means not using AI as a pretext for staff reductions whose real causes are weak financial results or poor planning. Employees sense this difference very clearly—and anyone who uses AI as an excuse squanders trust that cannot simply be bought back later.
Finally, there are employees who reject AI not out of fear but out of conviction—because of its energy consumption, unresolved copyright issues or a fundamental sense of unease. In the United States, requests from employees seeking exemption from AI use on ethical grounds are already becoming more common. One may agree with this position or not—the leadership question remains the same: how do I lead someone who does not want to use a tool for reasons of conscience when the company has committed to it? Anyone who reflexively dismisses such convictions reveals, in a small but telling way, how seriously they take their own values. This is not a peripheral issue. It is a litmus test.
What does genuine AI competence mean for employees?
AI literacy goes beyond simply knowing how to operate AI tools or craft effective prompts. Crucial to it is the ability to contextualize results, identify errors, question recommendations, and know when to trust one’s own judgment over the machine. This capacity for judgment is not developed in a one-off workshop but through practice, feedback, and experience.
Four fault lines, one common denominator: a half-day prompting workshop is not enough. The widespread notion that booking a training course means the workforce has been enabled fundamentally underestimates what is at stake.
True AI competence is not operational knowledge. It is judgment—the ability to assess when AI can be trusted and when it cannot. A particularly revealing finding comes from Aalto University. In an experiment, participants were asked to solve reasoning tasks with and without AI. One might expect those with more AI experience to assess themselves more realistically. The opposite was true: those who considered themselves competent in using AI overestimated their own performance more, not less. Superficial familiarity with the tool does not create sound judgment—it creates overconfidence. This is the real reason why a short workshop may not merely provide little benefit, but can cause harm: it instills precisely the confidence that increases the risk. Admittedly, this was measured in a laboratory rather than a boardroom—but the parallel is worryingly close.
What is needed instead is not a matter of operating tools, but of mindset—and this is where the real leadership lever lies. Whether a team atrophies or grows through AI does not depend on AI itself, but on how it is used. Convenient, thoughtless use—question in, answer out, accept it—leads to atrophy. Reflective use—questioning, cross-checking and comparing it with one’s own judgment—leads to growth. This distinction is not technical. It is cultural, and shaping culture is a core leadership responsibility.
Enablement therefore means routine, not a certificate. Knowing when to follow an AI recommendation and when to override it can be trained—but it requires practice, feedback and time, just as a professional tool cannot be mastered in a single afternoon. Leaders who genuinely want to make this possible for their people must create an environment in which “I did not trust the AI, and I was right” is recognized as competence, not obstruction.
Leadership is decided beneath the surface
Implementing AI is not a communication task. It is a leadership task at fault lines that remain invisible when one looks only at the rollout plan. The workforce splits. Work loses its core. The team stops asking questions. And some of the fear is justified. At each of these points, a leader can shape what happens—or announce that everything will be fine and hope that it is true.
It is the same question that has run through this series from the beginning, only on a different level: who is at the helm? This time, it is not about data or costs, but about cohesion and the development of one’s own people. This is the most demanding management task of all because it cannot be automated or delegated.
What is your view on this issue? Which of the four fault lines is your company currently facing—and are you talking about it openly, or is it still being brushed aside with a smile? I look forward to hearing your thoughts.
What to expect in this series
This series guides you through the key leadership issues surrounding AI in the enterprise—drawing on practical insights and firsthand experience.
So far, the focus has been on the people in the team. In the next article, I will turn my attention to leaders themselves—and to a danger more serious than being replaced by AI: the gradual erosion of their own judgment. Because anyone who wants to teach others not to trust AI blindly must begin with themselves.
In the previous article, I explored the question: Leadership in the Age of AI: Which Tasks Will Disappear—and Which New Ones Will Emerge?
Who this series is relevant for
This series is aimed at managing directors, board members, advisory board members, and executives who do not wish to reduce the approach to AI to a purely IT-related matter. It is particularly relevant for companies where AI is already being used in isolated areas, yet roles, responsibilities, and guardrails remain undefined.
Especially in medium-sized enterprises, this quickly creates a complex mix of opportunities, uncertainty, and pressure to act. That is precisely where leadership is required—not later, but now.
About the Author
He supports executive boards, management boards, advisory boards, and leadership teams in consciously shaping their roles, effectively leading change, and positioning their organizations for the future amid the challenges of digitalization, artificial intelligence, and regulation. His approach combines entrepreneurial practice, strategic thinking, and personal reflection—with a clear focus on sustainable impact.
He guides management teams, executive boards, advisory boards, and leadership teams in consciously shaping their roles, effectively leading change, and positioning their organizations for the future amidst the complex interplay of digitalization, artificial intelligence, and regulation. His approach combines business practice, strategic thinking, and personal reflection—with a clear focus on sustainable impact.
Find out more about our services Executive Advisory und Leadership Development. Get in touch – Contact
