KI-Governance within the company | Article 3
Few technologies are currently discussed in such reassuring terms as artificial intelligence when it comes to leadership. The narrative goes like this: AI relieves us of tedious routine tasks and finally gives us time for what truly matters. This narrative is true. And that is precisely why it is dangerous—because it obscures the real question.
In the last article, I described how AI is finding its way into companies in an unplanned and unmanaged manner—and why this is a leadership issue, not an IT issue. Today, we are going a level deeper: not looking at the technology within the company, but at the leadership role itself. What does AI actually do to what a leader does all day long?
Most executives I speak to ask the same question: “Which of my tasks will AI take over?” That is an understandable question. But it is not the crucial one. The more honest question is: What remains when AI can do everything that can be automated—and is what remains still leadership?
THE Truth LIES NOT IN THE PROFESSION, BUT IN THE TASK
The debate about AI and work suffers from a fundamental misconception. It asks which professions will disappear. That is the wrong unit of analysis. Professions rarely vanish wholesale. What disappears are individual tasks—and every profession is a bundle of many such tasks.
For leadership, this means: The question is not whether there will still be managing directors and board members in the future. Of course there will be. The question is which of the tasks that currently fill the daily lives of leaders will shift—and where they will shift to.
Let’s take a closer look. A significant portion of how executives spend their days involves information-related tasks: gathering status updates, reading and condensing reports, processing figures, and compiling materials for decision-making. It is about gaining the overview necessary to make decisions in the first place. AI can already handle much of this work today—faster, more thoroughly, and without fatigue.
What remains when this layer is stripped away? What remains is the very thing for which all that processing was merely a preliminary stage: orienting, prioritizing, and judging. Deciding what actually matters. Giving an organization direction amidst uncertainty. That is the essence of leadership—and that essence is not diminished by AI. It is laid bare.
THE END OF A COMFORTABLE HIDING PLACE
This is where things get uncomfortable. For many executives, operational work was not merely a burden—it was also a place to hide.
Anyone who filled their day with status updates, reporting, and forwarding information felt busy. Fully occupied. Needed. You could put in a twelve-hour day and go home feeling like you had led—even though, in reality, you had mostly just managed. The operational frenzy provided immediate proof of your own importance.
AI eliminates precisely that place of refuge. When the machine takes over the processing work, the time previously filled with staying busy remains—and with it, an uncomfortable visibility. It reveals whether someone uses that time for something only a leader can do, or whether nothing fills the void left by the administrative tasks.
Viewed in this light, AI is not merely a tool that makes a leader’s work easier; it is an X-ray machine. As tasks that can be automated disappear, it becomes ruthlessly apparent whether someone is truly leading—or whether the role was previously defined primarily by operational busyness. For some, this is a liberation; for others, a threat. For everyone, it is a question that can no longer be avoided.
So much for the shift within the role itself. Yet this does not happen in a vacuum. It encounters two developments that many leaders are currently underestimating—and which I consider the most dangerous stumbling blocks on this path.
STUMBLING BLOCK ONE: CUTTING JUNIOR-LEVEL FUNDING UNDERMINES YOUR OWN FOUNDATION
In many companies, I am currently observing a line of reasoning that seems compelling at first glance. The tasks with which entry-level employees traditionally start—conducting research, writing initial drafts, performing standard analyses, providing support—are precisely the tasks that AI handles well today. The conclusion, therefore, is that fewer entry-level staff are needed. Why pay a junior employee when AI can deliver the same results faster?
In the short term, the calculation works out. But in the long term, it ruins the company.
After all, no one is born with seasoned judgment. The judgment we rely on in senior staff is the product of years spent performing those very same small tasks—and making the minor mistakes that foster learning. Anyone who eliminates entry-level roles saves more than just a few salaries; they cut off the pipeline that is supposed to produce the senior professionals of the future. They saw off the very branch where their company’s seasoned judgment ought to be resting ten years down the line.
There is also a factor that is often overlooked. To judge whether an AI-generated result is sound, one must master the subject matter itself. There is a simple rule of thumb for this: you need to understand about eighty percent of a field yourself to recognize where the AI is going wrong. Anyone who has never learned the craft from the ground up cannot verify the work at the higher level; they can only rubber-stamp whatever the machine outputs. Consequently, an organization that replaces entry-level staff with AI does not merely create a shortage of new talent; it produces a generation of reviewers who are no longer capable of reviewing.
Interestingly, economics itself might correct this trend. Some AI experts are already pointing out that rising AI costs could once again make it worthwhile to hire junior staff. Which brings us to the second stumbling block.
STUMBLING BLOCK TWO: THE BILL ARRIVES – AND IT COMES FROM SOMEONE ELSE
Leadership is not just about leading people; it also entails setting priorities, taking responsibility for investments, and managing dependencies. This is precisely where the second pitfall lies—and it is vastly underestimated because it is often disguised as a technical cost issue, leading it to be conveniently offloaded to IT or the controlling department. That is a mistake. Deciding which technology a company relies on—and the extent to which it becomes dependent on a specific vendor—is a strategic leadership decision, not one that should rest solely with IT or controlling.
One of the most persistent assumptions about AI is that it is cheaper than humans. In 2026, many companies are discovering a catch to this assumption.
Since major providers switched their billing from flat rates to usage-based models, customers pay for every token processed—every single, minute unit of information. And bills are skyrocketing. The research firm Citrini Research has coined an apt term for this: “Token Panic.” At the IT service provider Adesso, token consumption increased nearly a hundredfold within just a few months, with costs reaching six-figure sums. Ride-hailing service Uber exhausted its annual AI budget in a matter of months—for a benefit that its own management openly describes as unclear.
An image circulating in developer circles hits the nail on the head: a man lighting a cigar with a flamethrower. The caption reads: “How I delete a file using the most powerful model.” It captures an entire mindset—using the most powerful, expensive tool for the most trivial task because no one is watching the costs.
And that is precisely the point: most do not look. A study by KPMG shows that only 26 percent of companies have a comprehensive overview of their AI costs. 22 percent only find out the actual expenditure when the bill arrives. It is the same blindness I have already described in this series—this time not regarding data, but money.
The smart approach looks different—and it is unspectacular. It begins not with IT, but with a leadership mindset: gaining clarity on where AI truly creates value and where it merely incurs high costs. Before applying AI to a process, you optimize the process itself. Anyone who layers artificial intelligence onto established yet inefficient workflows is essentially automating their own disarray—just at a higher cost. Streamline the process first, then introduce the machine. A real-world example illustrates the potential: when AI systems are provided with the right business context, token usage can be cut by more than half while reliability improves. Then there is the simple discipline of selecting the right model for the task rather than reflexively opting for the most powerful one—not every query requires a flamethrower. Furthermore, when costs or sensitive data are involved, smaller, open-source, and locally hosted models are gaining importance—sometimes fine-tuned with a company’s own proprietary data.
Underlying all this is a strategic question that extends far beyond the monthly bill. Major providers have invested billions in training and infrastructure. Market observers state the implication matter-of-factly: eventually, customers will have to foot the bill. Anyone who thoughtlessly aligns their entire business with just a few large models today may be optimizing operations in the short term—but is also locking themselves into a long-term dependency where they do not dictate the price.
This is the fundamental question of control for this series, applied to the company’s technological and financial sovereignty—and it belongs on the executive management’s agenda, not merely as a line item in a department’s budget. Who is holding the helm here—you or your supplier?
WHAT REMAINS – AND NO ONE TAKES OFF YOUR HANDS
Let us return to the initial question. What remains when the “can-do” tasks disappear and the two stumbling blocks have been avoided? Three things that no machine can take over. And it is no coincidence that these are precisely the three things that used to be frequently buried beneath the hustle and bustle of operational activity.
The first aspect is guiding people through the change itself. AI creates uncertainty, anxiety, and sometimes quiet resistance. It alters how teams collaborate and whom they trust. Guiding people through this transformation—providing them with direction where technology merely causes unrest—is a deeply human task. This will be the subject of the next article.
The second aspect is one’s own judgment. The more we delegate the preparation for decisions to AI, the greater the temptation to surrender the act of judgment itself. This is convenient—but it comes at a price that only becomes apparent later on. How a leader preserves their own capacity for judgment, rather than gradually losing it, is a serious question in its own right—one I will address separately.
The third element is responsibility. AI can prepare a decision, make a recommendation, and in many cases even provide a better rationale for it than a human can. What it cannot do is take responsibility for it. Responsibility cannot be delegated to a machine—and that is precisely what makes leadership in the age of AI not superfluous, but indispensable. That, too, deserves an article of its own.
AI DOES NOT MAKE LEADERSHIP SUPERFLUOUS – IT MAKES IT VISIBLE
The widespread concern that AI might replace leaders misses the point. AI does not replace leadership; it strips it bare. It removes the elements that allowed leadership to be simulated for years—operational busyness, information management, the mere appearance of being occupied—and lays bare what lies beneath. In some cases, what is revealed is a clear, guiding leadership personality ready to shoulder responsibility. In others, there is less underneath than the position suggested.
That is the real message of this shift. Not: Will I be replaced? But rather: What emerges when the administrative work falls away?
The good news is the same as at the beginning of this series: You don’t have to become an AI expert. But you do have to decide whether to use the time and attention you gain for what only you can do—or leave it to chance.
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.
The next article focuses on the first of the three remaining tasks: leading people through the AI transformation. After all, the greatest hurdle in adopting AI is rarely the technology itself; rather, it is the fears, the skills gaps, and the question of who ultimately bears the responsibility.
In the previous article, I addressed the question…: “Shadow AI in the Enterprise: Why Unplanned AI Use Becomes a Leadership Trap
But first, I am interested in your perspective: If you look honestly at your own day-to-day work as a leader—how much of it is administration, and how much is leadership? And what would become apparent if the administrative tasks were removed? I look forward to the exchange.
WHO THIS SERIES IS RELEVANT FOR
This series is aimed at CEOs, board members, advisory board members, and executives who do not want to reduce the handling of AI to a purely IT issue. It is particularly relevant for companies where AI is already being used in specific areas without clearly defined roles, responsibilities, and guidelines.
Especially in medium-sized companies, this quickly creates a complex mix of opportunities, uncertainty, and pressure to act. That is exactly where leadership is needed—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.
