AI transformation is not achieved through new tools, but through a new approach to work, responsibility, and collaboration.
Rolling out an AI license isn't a transformation in itself. It's a purchase.
What we see time and again in our work with companies is that many expect new tools to boost productivity—while overlooking the real problem. A tool doesn’t change collaboration or accountability; it merely accelerates what’s already there. Those who digitize old processes instead of rethinking them achieve efficiency without real progress.
The key question, therefore, is not: Where can we use AI? But rather: How do we need to redesign work so that AI actually creates value?
This could become uncomfortable. After all, AI doesn’t just change processes—it also changes influence, roles, and responsibilities. It raises questions that some would rather avoid: Which tasks will be eliminated? Whose expertise will become less important? How will decisions be made differently in the future? Leaders must accept that established processes aren’t automatically sound just because they’ve worked for a long time.
AI transformation is not an IT initiative that requires communication. It is a leadership task supported by technology. AI can analyze and make suggestions—but it cannot set goals, take responsibility, or provide direction. Those who merely implement tools end up with new software. Those who rethink work drive change.
AI provides answers every second. The problem is that not every answer helps a team move forward. Some are correct but irrelevant. Others seem convincing but are based on false assumptions. Still others provide a perfect solution to the wrong problem. The scarce resource of the future, therefore, is not information. It is judgment.
Leaders no longer need to have the greatest expertise in every area. Their job is to ensure that the team tackles the right problems, critically evaluates results, and remains capable of taking action despite a growing number of possibilities. This changes leadership in three key ways.
Anyone who defines leadership primarily in terms of answers suddenly finds themselves competing with a system that never sleeps and rarely says, “I’d have to think about that first.” The new strength lies not in giving the first answer—but in improving the quality of the question.
Not: How can we speed up this process? But rather: Why does this process still exist in this form at all?
Not: Which option does the system recommend? But rather: What criteria were used to evaluate this option—and what interests are missing?
Good questions draw attention, reveal assumptions, and prevent teams from veering off in the wrong direction too quickly.
AI speeds up work—resulting in more designs, more options, and more decisions in less time. Anyone who insists on reviewing every result and monitoring every step becomes a bottleneck themselves. The problem then isn’t the technology’s performance, but the leadership structure. Leaders must therefore clearly define: Which decisions is the team allowed to make on its own? When is AI allowed to provide support—and when not? Who ultimately bears responsibility? Without this clarity, two typical reactions arise: Either teams act too cautiously and seek approval for everything—or they use AI liberally without thoroughly assessing risks and consequences. Both approaches slow things down. This clarity is not bureaucracy. It is the prerequisite for speed.
AI can generate results that appear professional and sound logical, yet are still incorrect. Obvious errors are easy to spot. More dangerous are answers that seem plausible enough not to be questioned further.
Managers therefore need a better understanding of how quality is created:
Expert knowledge remains important—but it is no longer enough. What becomes crucial are the abilities to contextualize knowledge, recognize contradictions, and even be skeptical of a convincing conclusion if the underlying basis is flawed.
AI does not automatically make teams better. It first improves the quality of the system in which it is used.
We see this very clearly in our daily work with leadership teams: Clear goals are achieved more quickly. Unclear goals lead to hasty action without results. Good collaboration becomes more productive. Silos simply work at cross-purposes more quickly. A strong feedback culture accelerates learning. A culture of error avoidance ensures that risks remain hidden. And AI does not compensate for unclear leadership—it simply amplifies it.
That is precisely why leadership remains essential. Not because leaders need to master every technology in detail, but because someone has to set the direction, establish standards, manage tensions, and take responsibility for decisions. AI can expand the range of options. It is up to leadership to decide which of these have a future.
“Just try out what works for you.” That’s perfectly fine as a starting point. But it’s not enough as a working model.
After all, when each team member uses AI differently, many good ideas emerge—but no common approach to work has yet been established. Some are already automating entire work processes, while others use AI only for initial drafts. Results are evaluated in different ways, experiences remain with individuals, and valuable insights are not systematically shared. The next step, therefore, is to turn individual experiments into a deliberate division of labor.
In this context, it’s worth focusing not on entire professions, but on individual tasks. AI rarely takes over an entire role—it provides support for specific activities: researching, sorting, formulating, analyzing, and developing alternatives. As a result, other tasks actually gain in importance: classification, relationship-building, creativity, and responsible decision-making.
The key question is not: Which jobs will AI take over?
Rather: Which aspects of our work can AI usefully take over, and where do humans play a more important role?
| Mode | Description | Requirement |
|---|---|---|
| AI Takes Over | AI handles clearly defined tasks largely on its own; humans focus on exceptions and quality control | A consistent process, clear criteria, measurable quality, and errors that can be corrected quickly |
| AI-powered | AI provides analyses, proposals, and options; people add context and make the decision | Will become the standard in many areas |
| Humans lead | Tasks are deliberately kept in human hands because relationships, responsibility, and context are crucial | Far-reaching consequences, issues of trust, ethical considerations, decisions requiring explanation |
AI can help prepare for a termination meeting. A human should conduct it. AI can develop strategic scenarios. People decide what direction a company will take. It is precisely in situations where it’s not just about the right answer but also about responsibility that the value of human leadership becomes apparent.
The more complex, far-reaching, and difficult to correct a task is, the more human judgment must be involved. Five criteria help with classification: How complex is the task? What would be the consequences of an incorrect result? Does the decision need to be verifiable later on? Does it affect trust or reputation? And can it be easily corrected—or is it difficult to reverse? The goal is not to restrict AI as much as possible. The goal is to use it where it provides the greatest benefit.
AI transformation isn't just generating excitement. It's also raising questions that many people don't voice aloud: Is my experience still needed? Will I lose influence? Will I be replaced?
These are valid questions. And they deserve honest answers—not just reassurances.
What we see time and again in executive coaching sessions: Resistance rarely stems from complacency or a fear of change. It arises because of a lack of clarity. Because of a lack of security. Because of a lack of the feeling that one’s own perspective matters. Executives who ignore or dismiss this lose the trust of their teams—often quietly, often unnoticed, but always with serious consequences.
Trust isn't built through communication campaigns. It's built through concrete actions:
Even the best technology can fail due to a lack of trust. Conversely, a team with strong trust can achieve extraordinary results even with imperfect tools. That’s why transformation doesn’t start with technology. It starts with conversation.
AI has long been in use at many companies—just not necessarily where strategies and guidelines might suggest. Employees use AI for writing, analysis, and meeting preparation. Some have already established effective routines, others are testing it cautiously, and still others avoid it altogether. This is creating a lot of activity—but no common direction yet.
As long as positive experiences remain limited to individuals, only those individuals benefit. Processes hardly change, quality disparities widen, and the same mistakes are repeated in multiple places. Transformation only begins when individual tips and tricks are transformed into shared knowledge.
A good prompt does not necessarily make for a good process. It does not address whether the task was chosen appropriately, what data may be used, how the result will be verified, or who is responsible for it. Even the best prompt cannot save a poor process. A reliable process clarifies exactly these points—and shifts the focus from the individual prompt to the quality of the work as a whole.
Common standards create reliability without stifling the spirit of discovery. The right balance: clear guidelines for safety-related issues and responsibilities—and deliberate leeway to test new possibilities. Not every experiment has to become standard practice right away. But every good experiment should have the chance to do so. Learning should not be separated from day-to-day work: a monthly AI retrospective, a short demo session, or a concise case study are often enough to make experiences visible and actionable.
Success isn’t measured by the number of applications. A team can use AI every day and still achieve barely any better results. What matters is: Are decisions becoming more well-founded? Is quality improving? Is there more time for learning and challenging tasks? Do employees find the new way of working helpful? AI doesn’t drive transformation simply because as many people as possible use it. It drives transformation when it enables the organization to work better as a whole.
AI can make recommendations, assess risks, and prepare decisions. It does not bear the consequences of a decision.
This is precisely where one of the most important leadership challenges of the AI transformation lies. As long as AI only generates initial drafts of text, this question seems manageable. Things get more complicated when systems evaluate job applications, prioritize customers, or provide recommendations for action in sensitive situations. This quickly gives rise to a new form of ambiguity: The AI suggested it, the department adopted it, the manager approved it—but no one really sees themselves as the originator of the decision. Responsibility is spread across many parties and thus risks disappearing altogether.
AI systems often present results with great linguistic confidence. This style can easily obscure the fact that even a convincing answer is based on assumptions, data, and established criteria. A system does not evaluate neutrally—it evaluates within the framework created by humans.
These questions determine the quality of a recommendation. That is why, in any AI-supported decision-making process, it should remain clear where the analysis ends and the decision begins.
Depending on the task, these responsibilities may fall to one or more people. The key point is that no one should have to assume that someone else is already taking care of it.
Human review is only valuable if people can actually challenge an AI result. In everyday life, this is precisely what can be difficult: A system has analyzed large amounts of data and provides a professionally prepared recommendation—and anyone who deviates from it must justify their assessment in greater detail than someone who simply agrees. This inadvertently creates a new pressure to justify one’s position, which weakens human judgment. Organizations should therefore consciously foster a culture in which AI results are treated as informed contributions—not as unassailable authorities. A strong organization doesn’t just ask why someone is deviating from the AI ; it also asks why someone is following it.
Governance should provide guidance, not hinder work. Too little governance leads to hidden risks and decisions whose origins are difficult to trace. Too much governance creates lengthy approval processes and prevents teams from further developing useful applications.
The solution lies in a risk-based framework: The greater the potential impact, the more rigorous the review, documentation, and human decision-making must be. Good governance doesn’t just specify what is prohibited; it clarifies what is permissible and under what conditions.
AI transformation isn’t decided in strategy presentations. It manifests itself in day-to-day work—in tasks, decisions, coordination, and quality issues. Teams don’t need a master plan for every technological development. They need a shared framework within which they can experiment meaningfully, evaluate their experiences, and take responsibility.
These five questions have proven to be particularly effective in our practice—because they don't start with the technology, but with the work itself.
Don’t start with the tool—start with the problem. Where are we wasting time unnecessarily today? Which tasks are draining our energy without creating enough value? Where would better support make a noticeable difference? The more clearly the problem is described, the easier it is to assess whether AI is truly the right solution—and not just the obvious one.
In many cases, AI handles individual steps, while humans set goals, provide context, and make decisions. The most effective division of labor rarely results from a simple either/or choice. It’s not just a matter of what’s technically possible—complexity, the consequences of errors, the impact on relationships, and accountability are also crucial. The best division of labor leverages the strengths of technology without replacing human judgment in the process.
Just because something is convincing doesn’t automatically mean it’s correct, relevant, or appropriate. Teams should work together to define how good results are measured: Which facts need to be verified? Which assumptions need to be made explicit? What tone is appropriate for the given context? Without such standards, every AI result is evaluated at the user’s discretion—which leads to inconsistent quality and makes collaboration more difficult. A shared understanding of quality makes the use of AI more reliable, regardless of which tool or prompt is used.
The more AI is integrated into processes, the more important clear lines of responsibility become. Who is responsible for the data set? Who reviews the results from a technical perspective? Who makes the final decision? When should an issue be escalated? Responsibility must not become blurred between the system, the business unit, and management. AI can prepare decisions—but it can neither bear the consequences nor explain why a decision was justifiable in a specific context. This responsibility remains with humans—and it must be organized in a transparent manner.
Individual experiments drive momentum. Collaborative learning drives progress. Teams should regularly evaluate which practices truly help, where quality improves, and where additional effort is required. Mistakes, limitations, and unexpected side effects must be openly discussed. Short, established routines are often sufficient: a monthly retrospective, a case study discussion, or a brief demonstration of a successful use case. What’s crucial is that lessons learned don’t remain with individuals—and that best practices are translated into shared processes.
Managers don’t need to be familiar with every new application or have an immediate answer to every technical question. Their role is broader: They clarify which problem needs to be solved. They shape the division of labor between people and technology. They ensure quality and accountability. And they make sure that individual experiences lead to collective learning.
Leadership is becoming less of a source of all answers—and more of a shaper of a system in which people and AI can work together effectively.
The key question, therefore, is not how quickly a company implements AI, but rather how well the work that AI makes possible is organized.
Start your coaching training this fall and learn more about the curriculum and program structure now.
