Twelve layers of hierarchy compressed to roughly half. Around twelve thousand positions gone, with the number of management positions cut by roughly 50%. A rulebook of more than a 1000 pages being thinned. Some 2000 small teams now owning their work on 90-day cycles, with budgets shifting to where the work is rather than where it sat last year. Employees reporting that problems which used to take six months now resolve in days.
This is Dynamic Shared Ownership (DSO), the operating model introduced at Bayer in January 2024 by CEO Bill Anderson. The company was under pressure on multiple fronts — the loss of patent exclusivities in pharmaceuticals, mounting US litigation over Roundup, around 34 billion euros of net financial debt, and what Anderson described as a hierarchical bureaucracy that was blocking progress. The diagnosis was direct: 1,362 pages of internal rules, twelve layers between decision and customer, six-month approval cycles for things that should take days. The remedy was as direct as the diagnosis. Eliminate management layers. Replace annual budgets with ninety-day cycles. Reframe leadership as enabling rather than controlling. Let small teams own their work.
The early reports are encouraging. Like countless others, I am genuinely aligned with what the effort sets out to address. The drag of accumulated procedure, the distance between people and customers, the slow strangle of multi-layer approval — these are real problems, and Bayer is the largest company yet to take a serious run at them.
With this article I am putting a prediction on the record now, before it can be tested. If it is right, the structural reason will be worth understanding. If it is wrong, the structural reason for that will be worth understanding too.
Here is the prediction.
The argument in one move
Some choices in an organisation affect what others do. That part is structural. How that happens varies. In most organisations, the mechanism is directive authority: people act on the choice because they are expected to, regardless of whether they can see why. Hierarchy is one way of organising who holds directive authority. It is not the same as directive authority itself.
What is missing in most organisations is a requirement that the reasoning behind a directive be made visible to those who must act on it. Without that requirement, the reasoning stays in the head of whoever issued the directive.
That combination — choices others must act on, made without showing the reasoning, with no structural requirement to show it — is what produces the dysfunction commonly blamed on hierarchy. It produces procedures whose original logic no longer holds. It selects for confidence over insight, because confidence is what authority rewards when reasoning is not the basis of the choice. It hides the cost of being wrong, because the reasoning was never on the table in the first place.
There are three pieces to that combination: directive authority, a choice made without showing the reasoning, and no structural requirement to show it. Remove the directive authority and the choice no longer affects what others do. Install a requirement that the reasoning be made visible, and others can examine it — the reasoning either holds up or it does not. Either move dissolves the combination.
DSO has reorganised the first piece. It has distributed directive authority across two thousand small teams instead of routing it through twelve layers. What it has not done is install the requirement that the reasoning behind such choices be made visible to the people who must act on them. The third piece is intact. Bayer says ninety-five per cent of decisions are made by those on the ground; by design, DSO has not adopted a written governance constitution or any decision protocol requiring the reasoning to be shown before others act.
The combination has been redistributed, not removed. Two thousand teams now hold what twelve layers used to hold: the ability to make choices others must act on without showing the reasoning behind them.
Where the first tensions are likely to show
The current transition is sustained by an exceptional concentration of attention. The CEO personally hosts the transformation webinars. A Chief Transformation Officer holds a role created for this work. The ninety-day cycle keeps forcing a reset. All of this is doing real work. None of it would happen without the people currently driving it.
When that attention drops off — through CEO succession, the next crisis pulling focus, or the new ways simply becoming routine — what happens next depends on the structure.
A few places where tension is likely to show first:
At the seams between teams. Some choices affect multiple teams at once — pharmaceutical pipelines, technical architecture, multi-team standards. Without a structural requirement to show the reasoning behind such choices, they get resolved by whoever can bring authority to bear, however informally. Authority re-emerges, dressed in new vocabulary.
At the boundary with the outside world. External regulators, large clients, and litigation counterparts are organised hierarchically and expect to deal with someone who can commit on Bayer’s behalf. The roughly five billion euros in litigation payouts projected for 2026 will not be resolved by two thousand self-organising teams. The capability to commit lives somewhere — and where it lives, directive authority lives with it. Without a requirement to show the reasoning, the old pattern returns at the boundary and spreads inward.
In specialist-dependent domains. Patient safety, regulatory affairs, manufacturing quality — these depend on people who have built up specific expertise over many years, and on those people staying in their roles long enough for what they know to be reliably accessible. When people switch teams every ninety days, that kind of stability does not naturally arise. Specialist expertise tends to disappear quietly; it is noticed when something breaks.
Inside the surviving management positions. Roughly half of Bayer’s managers remain, reframed as VACCs — visionaries, architects, catalysts, coaches — and asked to enable rather than control. They still hold directive authority. Under pressure, without a structural requirement that they show the reasoning behind their choices, the path of least resistance is to revert. The reframing names the aspiration. It does not change the structure they are in.
At CEO succession. DSO is the personal project of one CEO. Its roots — Gary Hamel’s Humanocracy, Haier’s RenDanHeYi — show the pattern: post-hierarchy models depend on a champion who keeps the architecture alive. Bayer’s first real test will be whether what has been built can survive without the person who built it.
What earlier post-hierarchy experiments suggest
Bayer is not the first organisation to attempt what DSO attempts. The pattern in earlier attempts is telling.
Zappos adopted Holacracy company-wide in 2014. After the CEO issued an ultimatum to commit to the new model or leave, around eighteen per cent of the workforce took the buyout. Over the following years, Zappos quietly walked back significant parts of the implementation — bringing back managers and scaling back the formal meeting structure, while keeping the circular team architecture. Holacracy provided elaborate governance procedures. What it did not require was that the reasoning behind such choices be made visible to the people who would have to act on them. Without that requirement, the new procedures became another layer of process to navigate, and the patterns Holacracy was meant to replace came back inside its own formal structure.
Medium, founded by Twitter co-founder Ev Williams, abandoned Holacracy in 2016. Its head of operations explained that the company had been spending too much time on procedural overhead and not enough on the work that mattered.
Beyond Holacracy specifically, GitHub was famously flat for years before re-introducing management positions as the company scaled. The harassment case that became public in 2014 showed what had been quietly true for some time: removing formal authority had not actually removed authority. It had moved into informal channels, where it was harder to see. Valve has been cited for two decades as a successful flat organisation; former employees describe an effective hidden hierarchy, with influence gathering around long-serving engineers and around personal ties to the founder.
Haier’s RenDanHeYi, the model Bayer explicitly cites as inspiration, has produced real and measurable outcomes across decades. It has also operated continuously with Zhang Ruimin championing it; whether the architecture survives his departure is, as of now, open.
The common thread is not that these models were wrong. They each addressed something real. They distributed or removed formal authority without installing the structural requirement that the reasoning behind such choices be shown to those who must act on them. Where they have held up, it has typically been through sustained championship from the top. Where that championship has weakened, the patterns the model was meant to replace have come back — often inside the model’s own formal structure.
The prediction, plainly
Large organisations rarely fail in dramatic public ways. They degrade quietly.
Several years from now, once the exceptional attention from the top has dropped off, the prediction is that three things may become visible to anyone looking closely.
Coordination friction between teams will have brought informal authority chains back.
At the boundary with the outside world — regulators, large clients, litigation — a parallel hierarchy will have grown, and the internal teams will quietly defer to it.
Specialist domains that the ninety-day marketplace cannot keep stable will either have built up risk or have quietly gone back to traditional staffing, outside the DSO model.
None of this would mean the experiment was not worth running, or that it did not conjure up something useful to the organisation. DSO has clearly reduced one form of organisational drag, and it is teaching many organisations to see how much built-up procedure was load-bearing on nothing. Those are real contributions. The structural argument is narrower: DSO has reorganised the first piece without touching the third.
Where this prediction comes from
The reading follows from a framework developed for the past decade — Decision Free Solutions (DFS). DFS rests on three foundations. First, that the world is causal — outcomes have causes. Second, that people vary in how much they perceive of what is in front of them. Third, that “decision” has a precise meaning — a choice made without showing the reasoning behind it. That third definition is what distinguishes choices that add unnecessary risk from those that do not. Everything else in the framework follows from these three starting points by definitions and logic. That is what makes DFS deductive rather than empirical — predictions like the one above come from reasoning, not from hunches or extrapolation. The argument above is what the framework predicts.
If the prediction holds, the next phase of the conversation Bayer has started will arrive on its own. If it does not — if Bayer demonstrates that distributing directive authority alone is enough, that no requirement to show reasoning is needed — then the framework has no organisational value, and the world will have learned something important about how organisations can work.
Either outcome is worth more than silence.
So this is on the record now. Let us see.
Epilogue (added 26 May 2026)
This is not a critique of DSO. The design move it represents addresses something real, and could yet hold. What stands between the prediction here and a different outcome is — following the DFS framework — a structural condition that can be added within DSO, not a different methodology to switch to.
A companion article will follow. Its working title is “2031: How Bayer managed to make DSO a success after all”. It will not be an afterthought to this one. It will be a structural piece — an imaginary scenario in which the conditions needed to make DSO succeed are put in place, and what that looks like in operation. The framework that produces the prediction here is the same framework that can sketch the alternative; the second article applies it constructively rather than predictively.
This article is first because the structural reading suggests it. The future companion-article is what makes the conversation symmetric.