A team visits a company famous for fast, clean decisions. They come home with the artifact: the one-page decision doc, the weekly cadence, the decision log, the escalation rule. Six months later every artifact exists and the decisions are still slow.
Nobody is doing it wrong. The doc is filled in. The meeting happens. The log is updated.
The system was never the doc.
What the system actually is
A decision system is a set of answers to four questions that almost never get written down:
- Who is allowed to decide, and who is allowed to stop a decision after it is made?
- What counts as evidence here?
- What does being publicly wrong cost?
- How fast does the ground move under us?
The visible artifact is the answer printed on paper. The working system is whether the ground agrees with the print.
| What you copied | The hidden assumption it rests on |
|---|---|
| One-page decision doc | Someone senior will honor a written objection from someone junior |
| Weekly decision meeting | A week is not too slow, and nobody above will pull the decision out of the room |
| Consensus circle | Stakes are shared enough that slow agreement beats fast command |
| Owner decides, others advise | Individual judgment is trusted more than group smoothing |
| Decision log | People read it, and history is allowed to embarrass |
| Escalation path | Escalating is career-safe, and someone above is actually listening |
A system works where those assumptions are true. The same system dies where they are not, and it dies quietly, because everyone keeps filling in the forms.
Why copies fail without noise
Think transplant, not installation.
A decision system grows in a specific soil: a resource base, a trust web, a history of who was right and who was punished, a pace of consequence. The visible artifacts are the leaves. The assumptions are the roots. A copy takes the leaves and leaves the roots behind.
The transplanted system does not usually crash. It performs. Meetings happen, docs get written, and the real decision moves somewhere else: the hallway, the private call, the loudest voice. The ritual continues because the ritual is cheap. The assumptions are what cost.
This is the same trap as copying the behavior without entering the life. The behavior is observable. The life conditions that produced the behavior are not.
The agentic version of the same mistake
Now the mistake has a new costume. Teams build decision systems for agents: contracts, gates, verification steps, escalation rules. And the old sentence comes back: this one will transfer, because it is consistent, because it is written down, because it is rational.
Agents change the speed of execution. They do not change the ground the contract grew from. The done-contract encodes someone's answer to all four questions: who may accept, what evidence counts, what a wrong accept costs, how fast the surface moves. Ship those contracts to a group living in different conditions and the agents will faithfully enforce assumptions the new ground does not hold.
Rational does not mean portable. Written down does not mean true.
What travels and what does not
| Travels | Does not travel |
|---|---|
| Formats and templates | Trust between specific people |
| Vocabulary for decisions | Appetite for public misses |
| Checklists and gates | Memory of past consequences |
| Meeting cadence | Real authority of the people in the room |
| Escalation diagrams | Whether escalation is actually safe |
The left column is what a visit shows you. The right column is what made the visit worth writing about. A system is portable exactly as far as its right column is already true on the receiving ground.
How to read the ground before you plant
Before importing any decision system, five questions. Ask them of the receiving ground, not of the system:
- Who actually stopped the last bad decision here, and what did it cost them?
- What happened to the last person who was publicly wrong?
- When the data runs out, what does this group do? Instinct, delay, or a senior voice?
- How much runway stands behind the word "no"?
- Which floor of the organization is this system for, and did anyone ask that floor?
If the answers contradict the system's assumptions, the system is decoration. You can still import the leaves: the template, the vocabulary, the checklist. Just know which column they belong to.
FAQ
Is the problem that teams copy systems too eagerly?
No. Copying is fine and often wise. The problem is copying the artifact and assuming the assumptions came with it. Name the four assumptions out loud, then check the ground. Most systems survive the transplant once you adjust them to the receiving conditions instead of pretending the conditions match.
Does this mean context beats design every time?
Context beats imported design. Designed systems do work when they are grown in place: built with the people who live in the conditions, priced against the resources actually there, and adjusted when the ground moves. The failure is not design. The failure is design done somewhere else.
Where do AI agents fit into this?
Agents make the stakes higher, not lower. An agent system scales whatever assumptions are encoded in its contracts, faster and more consistently than humans would. If the assumptions mismatch the ground, you get confident wrong output at machine speed. The transplant question comes first.
What should I read next in this series?
The next post explains levels of development without turning them into a scoreboard, because "different ground" needs a working vocabulary that does not become an insult system. After that, the series walks through floors, decision technologies, resource prices, communities, and judgment.
Related reading
- Copying the Behavior Is Not the Same as Entering the Life
- A Good Move From the Wrong Ground Still Costs You
- You Cannot Borrow Another Person's Timing
- Workflow Theater vs Workflow Gain
- The Proxy Becomes Dangerous When It Replaces the Problem
If you are deciding which decision system your team should actually live in, book a discovery call.