Most companies that scale revenue faster than they build operating architecture end up in the same place. Demand floods in through informal channels. The founders become the routing layer. One function quietly absorbs the incoherence. And the cost compounds, invisibly, because nobody is measuring it.
Usually the function that absorbs it is product. Deals stall in delivery. Sprints break. One approval circles the same meeting for months. The obvious read is a product problem, and the obvious move is a product hire.
The obvious read is usually wrong.
In our work the pattern resolves the same way almost every time. It is not a product problem. It is an operating-model problem that shows up most visibly through product. Product has become the shock absorber for every missing system: intake, prioritisation, decision rights, customer feedback, strategic translation. A new product leader adds capacity to the shock absorber. It does not fix the thing being absorbed.
Why it stays invisible
Operating-model incoherence has no line item. It surfaces as rework, delay, duplicated effort and executive time spent firefighting. Everyone feels it. Nobody can point to it. So it never gets prioritised against things that carry a number.
We give it a number. We call it the coherence tax: the annual cost an organisation pays for operating below its coherence potential. In one organisation we instrumented, the reconciled coherence tax ran into eight figures a year. Evidence-based, not accounting-grade, with confidence stated per pattern. Once a cost like that is on the page, the conversation changes.
Measure, do not opine
Most operating-model reviews produce a slide of themes. We produce a baseline. Our Qualitative Coherence Indicators score how an organisation actually operates across six dimensions: strategic, governance, structural, role, cultural and operational coherence. We derive it twice, once from coded qualitative evidence and once independently from a whole-of-staff instrument, and we check that the two converge.
Two findings from that work matter more than the rest.
People are almost never the problem. In the organisation above, psychological safety was the highest-scoring item in the survey while trust in systems sat well below it. People did not blame each other. They blamed the architecture. That is why the whole-of-staff instrument reached full completion, and why not one person interviewed was assessed as resistant to change. Blame the system rather than the people, and the people help you fix the system.
The capability to operate coherently usually already exists in-house. One subsystem in that same organisation, run with clear ownership and low interference, scored more than double the company baseline on the same scale. The floor is not a talent problem. It is a design problem.
Prediction, then observation
A diagnosis is a set of claims about what is wrong. The honest test is whether those claims come true.
Our sensing function ran live for ten weeks against the engagement's real signal streams. Near-daily. It produced a dated archive of 113 reports. The friction patterns we had named in the diagnostic were then caught firing in real time, on dated documentary evidence. A deal priced, scoped and committed before discovery was complete, flagged the same afternoon. A senior request approved in principle that then went unactioned for days, named as the accountability pattern it was. Prediction, then observation.
The instrument also reported against its own operator. When our own unsent drafts began enacting the very pattern we had been engaged to detect, the briefings said so, repeatedly and on the record. An instrument that indicts its own side is an instrument whose favourable readings mean something.
The uncomfortable part
Detection is not correction. Ten weeks of accurate sensing also proved the harder point: sensing alone does not move an organisation. The register went stale and the work stalled while the monitor watched it happen. That is not a weakness in the evidence. It is the strongest argument for the rest of the system. Eyes tell you where the leak is. They do not close it. Routing, decision persistence and follow-through do.
Why this matters now
The AI transition sharpens all of it. As organisations deploy agents and automated workflows, the binding constraint stops being technical capability and becomes organisational coherence: whether signals route, whether decisions persist, whether authority is legible enough for autonomous systems to be governable at all. You cannot delegate to agents inside an operating model your own people cannot navigate.
The layer that makes AI land is the layer above agent governance. Most organisations have not built it. That layer is what we build, and we measure whether it holds.