When the same conflict survives everyone who leaves
The Architect stage
Three teams have been fighting for two years. Two of the original protagonists have left and been replaced, and within four months the new people were having the same fight in almost the same words.
If your instinct there is to ask what would have to be true about this organisation for all three teams to be behaving completely reasonably, you are probably leading from the Architect stage.
The conflict is not a relationship failure. It is the organisation’s own design, expressed through whichever humans happen to be occupying the roles.
What it looks like
Behaviour is largely generated by structure — by what gets measured, what gets rewarded, who has access to what information, where the decision rights sit, what is easy and what is hard. Change the structure and the behaviour changes, generally without anybody being told anything.
You can hold competing legitimate interests without collapsing them: seeing that sales and engineering and finance each have a real case, that the tension is structural rather than a misunderstanding, and that the job is to design something that holds it productively rather than to declare a winner.
You are comfortable with long feedback loops — making a change and expecting to know whether it worked in nine months, which requires a tolerance for extended ambiguity that most organisations actively punish. And you have mostly stopped needing credit, partly through development and partly through necessity: when your interventions are three steps upstream, nobody can trace the result back to you anyway.
Where it goes wrong
The characteristic failure mirrors the Cultivator’s. Having discovered that structure produces behaviour, you can start treating everything as structural — so an individual who is genuinely the problem gets a redesigned process instead of a conversation. Sometimes the incentives are fine and someone is simply behaving badly, and this frame is bad at that, because its whole insight is that blaming individuals is naive.
Distance. Systems work is done at a remove, and the remove is easy to extend. A year on operating models and governance without a hard conversation with an actual person can be justified as working at the right level. Some of it is. Some of it is the same flight into competence the Practitioner makes, at higher altitude.
Impatience. Structural intervention is slow and organisations are not. Architects frequently find their most important work cancelled at month seven of a nine-month feedback loop, by someone who wants to see something.
What moves you from here
You can shape a system. The next question is what the system is for, over what horizon, and who counts.
It tends to arrive through a specific discomfort: you design something that genuinely improves the organisation, then notice the cost has landed somewhere outside the boundary you were optimising within — a supplier, a community, a group who were not in the room. The system got better. Something outside it got worse, and the system had no way to register that.
Or it arrives through time. You look at something you built five years ago and find it degraded, worked around or quietly reversed, and have to sit with what would have made it hold.