The previous articles in this series treated coordination, commitment, and learning as three separate capabilities. That separation was useful for diagnosis. But it is misleading as a description of how organizations actually function — because the three are not parallel capabilities that can be improved independently. They are sequential phases of the same underlying process. An organization that can run all three phases at low cost and high fidelity is in a categorically different position from one that cannot. That difference is the subject of this article.
The Three Phases
Every organization that survives over time runs some version of this cycle.
Coordinate — bring distributed knowledge into alignment around a shared model of the situation. This is the work that happens before a commitment: gathering signals, interpreting them from multiple vantage points, surfacing disagreement, integrating evidence into a picture clear enough to act on. It is the Orient phase of the decision loop, played out across many people who each hold a different piece of the relevant information.
Commit — bind the collective to a course of action. Allocate resources. Create obligations. Make promises that outlast any individual's presence. A commitment is not a plan or an intention — it is an act that changes what the organization owes, what it can and cannot do, and how it will be held accountable. Commitment without sufficient coordination is fragile: the assumptions behind it are untested, the alignment behind it is shallow, and the first contact with reality tends to shatter it.
Update Priors — revise the shared model based on what happened versus what was expected. Integrate what was discovered into the starting point for the next cycle. This is not post-mortem analysis or lessons-learned documentation. It is the act of changing the organization's priors — the working models and assumptions that will govern how the next situation is read, how the next commitment is formed, and what evidence will be sought or ignored. The endpoint is not the lesson. It is the updated shared model. The adaptive advantage — the compounding — comes from what the next cycle starts with, not from what was recorded about the last one.
These three phases are sequential. Each depends on the previous one. And crucially, each feeds the next cycle: updating priors from this commitment improves the quality of coordination in the next one, which improves the quality of the commitment after that, which produces richer learning from the one after that.
What Breaks When One Phase Fails
The failure modes are distinct and recognizable.
Coordination without commitment produces deliberative organizations that cannot act. The best-informed, most thoroughly discussed decision is perpetually deferred because the cost of committing feels higher than the cost of continuing to deliberate. Uncertainty is used as a reason to avoid commitment rather than as a condition to manage it.
Commitment without coordination produces fragile organizations that act badly. Decisions are made quickly, by the right authority, with the wrong shared understanding. Sales commits to a delivery date operations never agreed to. Leadership commits to a strategy that the front line does not believe is executable. The commitment is made, but the alignment that would make it keepable was never established.
Commitment without updating priors produces efficient organizations that stagnate. They execute well. They keep their promises. But they do not improve. The same types of decisions are made with the same quality of shared understanding, over and over, because the organization's beliefs are not being updated by the outcomes it produces. The cycle runs, but it does not compound.
Updating priors without coordination produces disconnected organizations that cannot apply what they know. Each part learns from its own experience. The engineering team learns from its failures; the sales team learns from its wins; the leadership team learns from its board conversations. None of this learning propagates into the shared model that governs the next commitment. The organization is rich in distributed insight and poor in collective wisdom.
Why the Cycle Is Usually Slow
The adaptive cycle has always existed. What varies between organizations — and between eras — is how long each phase takes, how much is lost in the transitions between phases, and how much of each cycle's output makes it into the next one.
In most organizations, today:
Coordination is expensive. The work of bringing distributed knowledge into alignment requires meetings, documents, presentations, and lengthy email chains. Much of the relevant information never surfaces. What surfaces is filtered and translated repeatedly before reaching the people who need it.
Commitment is difficult. The threshold for commitment is often set by the organizational dynamics of the room rather than by the quality of the evidence — authority, confidence, and social relationships doing more work than the actual state of knowledge.
Updating priors is sporadic. Post-mortems are held when something goes dramatically wrong, not routinely after every significant commitment. What is retained is the artifact — the decision record, the action log — not the belief that made the decision sensible or fragile. The next cycle starts from approximately the same prior as the last one.
The result is an adaptive cycle that is slow, lossy, and largely non-compounding. Organizations improve operationally as people gain experience, but they do not improve cognitively as organizations.
What Happens When the Cycle Accelerates
An organization that can run the adaptive cycle at significantly lower cost and higher fidelity does not simply become more productive. It becomes more adaptive — capable of changing faster than its environment changes, rather than being changed by its environment.
The compounding effect is real and significant. Each completed cycle that retains genuine learning — updating beliefs, not just logging outcomes — starts the next cycle from a more accurate prior. The coordination in the next cycle is therefore more efficient, because the shared model it draws on is already closer to the current reality. The commitment is better formed, because it is based on beliefs that have been calibrated by experience. The learning is richer, because the prior was explicit enough to compare against what actually happened.
The organization does not just run faster. It runs better, in proportion to the quality of what it retains from each cycle. This is the compounding story: not that individual decisions improve, but that the starting point for each decision is systematically better than the last.
The Frontier Question
Each technological era changed the cost of running the adaptive cycle. The telephone reduced the cost of coordination. The database reduced the cost of commitment tracking. Email reduced the latency between phases. Each change mattered, and each was adopted into organizational structure.
The current transition is different in kind. For the first time, information technology can participate directly in all three phases simultaneously — not just transporting information between them, but actively reducing friction within each.
In coordination: continuously synthesizing distributed evidence into a shared model, surfacing what is known and what is contested, without requiring the information to travel through human translation chains.
In commitment: supporting the formation of commitments with structured evidence, explicit assumptions, and clear stopping criteria — and preserving the belief state that justified each commitment alongside the commitment itself.
In updating priors: calibrating the organization's working model of its situation continuously from each completed cycle, rather than reconstructing from scratch each time a similar situation arises.
AI is the most significant instance of this shift so far. The question is not whether this changes the adaptive cycle. It already does, for organizations that have adopted AI-assisted collaboration seriously. The question is whether organizations will build the infrastructure to make that change systematic — or continue to treat it as an individual productivity tool that touches each person separately without ever improving the cycle as a whole.
That is the organizational design question of the current moment. And it is not a question about AI. It is a question about what organizations need to remain capable of learning — and what it costs them when they cannot.
Core thesis: Coordination, commitment, and updating priors are one cycle. Breaking any phase doesn't just lose that cycle's learning — it loses the compounding of every cycle that follows. The idea you can't unsee: Organizations don't just run the cycle faster when technology improves. They run it more faithfully — which is what produces the compounding. Vocabulary shift: "we need better processes" → "we need to close the adaptive loop" Connects to: Article 1 (The Organization as an Adaptive System), Article 2 (The Physics of Organizations), Article 4 (The Commitment Machine) Version: 1.0 / 2026-07-06