The informational asymmetry on which classical corporate leadership rests is tipping. Employees are rebuilding workflows their superiors no longer grasp. Those who do not get moving themselves lose the power of interpretation — not curable by training or external consultants.
Why your team is overtaking you right now
Essay #001 · 2026 · Reading time 10 min · Field: Sovereignty · Market: all
There is a moment in which most entrepreneurs notice that something has tipped. It does not come with a fanfare. It arrives on a Tuesday afternoon, when an employee says something in a meeting that is technically more precise than anything you yourself had thought about the subject. You nod. You make a note. Later you find yourself asking whether you still actually know what is happening in this company.
That is the moment in which arocon begins.
Most entrepreneurs have a good sense of where their personal knowledge carries and where it no longer does. They know the market, they know the numbers, they know how a customer thinks, and they know which conversations they must have and which they need not. That security is not a coincidence. It was built over years.
But in the past two years, something has been added that selectively overrides this security. A topic that, unlike earlier ones, cannot be caught up on across two weekends of reading. A topic that opens new layers within itself the moment you think you have grasped one. A topic whose knowledge is rapidly unevenly distributed inside your own workforce — and not along the hierarchy.
The marketing assistant who has been experimenting with agentic systems for ten months knows more of their mechanics than the head of marketing above her. The young controller who runs models locally in his spare time formulates more precisely than the CFO. The developer in IT builds an internal co-pilot that outperforms the official solution after a week.
These are not isolated cases. This is the new state.
What agentic actually means
Before I explain why this is a structural problem, a short aside on what distinguishes agentic work from everything else that has entered the company as a digital tool over the last fifteen years.
A classical tool — a CRM, a DMS, an ERP — is a container in which humans do work. The human knows what he wants, and the tool helps him do it efficiently. The authority remains with the human.
Agentic systems invert this. They are not containers but actors. They receive a goal, not a step of work. They decide themselves which steps to take. They ask when something is missing and deliver a result that is often already usable. In more complex arrangements, several agents collaborate — one researches, one structures, one formulates, one checks.
The consequence: work that used to take hours or days of a human’s time gets done in minutes. Not perfect, but good enough in seventy or eighty percent of cases. And the twenty or thirty percent of rework is exactly that — rework, not foundational work.
That sounds like an efficiency story. But it is only efficiency on the surface.
Under the surface lies a shift in where operational knowledge is stored. Employees used to know how their work is done, and the entrepreneur could ask them how it is done. Today, the agentic system knows how the work can be done — and the human begins negotiating with it. Anyone who does not understand how the system structures the work no longer understands where the critical points sit.
Hierarchy rests on asymmetry
Classical corporate leadership works because, in sum, the decision-maker knows more than any single person in the company. Not more than each individual specialist in their field — but more in aggregate. He sees connections no one else can reach. He joins single pieces that others hold in silos. This overview knowledge is the invisible foundation of authority.
This has worked for about two hundred years, ever since the modern form of the enterprise emerged. It works even though individuals in the company can be selectively ahead of their boss in their area. That selectivity of getting ahead was the key: the head of sales was ahead in sales, the designer was ahead in design — but the entrepreneur had the overview that held the parts together. In sum, he knew more, even when he knew less about controlling than the controller.
The decision-maker’s authority has two layers. The upper one is formal: employment contract, company law, register entry, directive authority. The lower one is functional: “My boss knows what he is talking about.” The upper layer only carries when the lower one carries. When the lower one does not carry, a quiet, invisible loss of authority sets in — not as rebellion, but as decoupling. Employees listen politely, nod in agreement, and internally decide that they must do the real thinking themselves. Decisions continue to be formally signed by the boss. In substance, they are thought at other tables.
That is the state into which many German-speaking Mittelstand companies are sliding right now. Not dramatic, not loud, but distinct. And it does not go away on its own. Every month, the age of agents widens the gap a little, because the learning curve at the lower end of the company is steeper than at the top.
The paradox: it is in large part not the decision-maker’s fault. He did not suddenly stop being good. The world shifted while he continued with his accustomed mode of work. But precisely for that reason, left unaddressed, the problem is relentless.
Three symptoms
Three symptoms are easy to observe.
The first: proposals you can no longer judge sovereignly. A department arrives with a proposal that uses agentic systems — a customer-service bot, an internal knowledge assessment, a sales assistant. You get a presentation. You nod in the right places. At the end you either green-light it or turn it down — but you know you have not made the decision. You have delegated without wanting to delegate.
That is not shameful. It is a signal.
The second: investments whose assumptions you cannot penetrate. An IT lead proposes an infrastructure investment: vector database, embedding models, orchestration, monitoring. Sums in the mid six-figure range. The numbers are plausibly calculated. But you feel that the actual assumption — that all of this will still be relevant in three years — may be built on sand. The proposal itself cannot answer that. The proposal is written from the perspective of those who like the proposal.
In the classical logic you would now speak with your advisory board. But the board has usually not penetrated the topic any more deeply either.
The third: personnel decisions whose productivity math no one can reconstruct. A department in which agentic workflows are used intensively comes with a staffing question: does this role drop away, does that role grow, is an entirely new one needed? The manager has an opinion. You have another. The question of who is right cannot be decided by experience. Because experience for this constellation does not yet exist.
This is the point at which many entrepreneurs notice for the first time: I currently do not have an opinion that rests on my own knowledge. I have an opinion that rests on reflex.
That, too, is a symptom.
What does not work
There is a range of answers offered to this symptomatology. Most of them do not work.
AI training for the team. The team is not the problem. The team is part of the solution — at least in part. Those who respond with training treat the symptom, not the cause. And they give away their own time, which they would have needed in order to get moving themselves.
An external Chief AI Officer. A new role with AI responsibility is a construction that shifts the problem rather than solving it. The CAIO stands between you and the decisions you must make. He in turn becomes a source of informational asymmetry — now even with an official mandate. You no longer ask your developer, but an advisor who stands between you and her. That is not sovereignty. It is another delegation.
The strategy deck of a consultancy. A well-written deck is reassuring. It orders. It prioritizes. It gives recommendations. But it does not replace your own thinking-through. What you need as a decision-maker is not a deck — it is the ability to examine the deck yourself. Those who take a deck as an answer have not understood the question.
A broad tool rollout. Tools are downstream. Those who answer the question with a tool have shrunk the question. The question is not which co-pilot package you buy. The question is what role your organization plays in a world in which work is divided up differently. The question is a question of order, not of product.
All these answers share a common denominator. They try to solve the problem through external expertise. That is precisely the mechanism that produced the problem in the first place.
What does work
What works is uncomfortable. The decision-maker must learn himself.
Not to become a developer. Not to configure agents by hand. But to be able to ask the questions that cannot be delegated from outside.
Which questions are those? They are not technical. They are architectural.
- How does our value creation change when thirty percent of today’s activities are carried out agentically?
- Which competencies lose weight, which gain?
- Which roles must we replace, which re-cut, which create anew?
- What may agents decide, what not, and who checks that?
- Which investments are necessary, which are fashion, which are false security?
- What does liability mean when decisions are increasingly no longer made exclusively by humans?
These questions cannot be answered without the decision-maker himself setting a foot into the matter. Not into the console, but into the logic. Not into the configuration, but into the principle.
How much time does it cost? Less than most fear. A few hours per week, consistently invested across three to six months, puts a decision-maker in a position where he can ask the right questions, fairly judge the proposals of his people, and restore his own authority. Not by being smarter than his staff — but by seeing the architecture in which they work again.
Seeing the architecture means: recognizing where the decisive levers sit. Where the company actually creates its value, and how agents amplify, shift, or undermine that value. Those who see it no longer decide defensively. They decide strategically — from their own judgment, not from trust in external recommendations.
This is not a skills program. It is a posture that emerges through work. And it is the only answer that addresses the core of the problem instead of displacing it.
arocon accompanies this path. In confidential 1:1 sparring, in monthly essays, in small rounds among equals. No course programs, no certificates. A working relationship for those who know that they must get moving themselves.
Because the lead belongs to the decision-maker. Not to the team, not to the consultant. To the person who ultimately decides — and who refuses to do so from the second row.
— Axel Roth