The Fix

Chapter 16: The First Unit

The first autonomous unit is a permanent structural change, not a pilot. One unit operating with genuine autonomy creates selection pressure that shifts the question from “should we restructure?” to “why does everything else still work the old way?”

The unit lead sits down with the domain expert and two engineers to write the process definition for claims management. It is the first working session, and the whiteboard is clean.

Within thirty minutes, they find the first discrepancy. The domain expert describes a triage step: incoming claims are assessed for complexity and routed to the appropriate handler based on claim type, value, and regulatory category. The senior engineer pulls up the service, and the triage logic turns out to be a hard-coded routing table, last updated nineteen months ago, before a regulatory change that added two new claim categories. Those categories were applied in the code as exception handlers that bypass the routing table entirely, and neither person had seen the other's half: the domain expert had never seen the exception handlers, and the engineer had never seen the triage logic the domain expert describes.

Both of them are right. The documented process and the implemented process diverged after the regulatory change: the code was patched, the operations manual was not, and for nineteen months the claims team has been triaging the two new categories by hand, using a workaround that nobody outside the team knows about.

She says: “This has been wrong for nineteen months. We have been triaging manually.”

The engineer says: “I didn't know there was a manual process.”

Together they write the first line of the process definition, and it describes what actually happens.

The full definition the unit eventually produced, every state, transition, failure mode, and contract dependency, appears in the appendix. The first line took thirty minutes; the full definition took three weeks. The first autonomous unit holds real authority over its process, not a sandboxed experiment the organisation can quietly walk back. It operates under its own governance charter and is evaluated by outcomes rather than by a committee. Its job is not to guarantee a transformation but to produce evidence that cannot be outlasted the way a memo can: one unit operating with genuine autonomy creates selection pressure, making the cost of the surrounding dysfunction visible by contrast, so the question “should we restructure?” becomes harder to keep asking beside a process that already has.

What the unit changes is narrower than a forcing function and more durable than an experiment. The organisation's most reliable instrument against inconvenient truth has never been argument but sympathetic neglect, the slow withdrawal of attention that lets a dissenting engineer, an unwelcome finding, or a platform business case fade without anyone having to be seen rejecting it. A unit that reconciles its process against the running system every week, and publishes the contracts it lives by, cannot be let go quietly in that way. Its results carry dates, its decline is measurable, and ending it becomes a decision with an author rather than a thing that simply stopped being mentioned. That does not stop an organisation determined to reclaim the territory, and the determined ones still can; what it removes is the cheapest route, the one that requires no one to decide anything.

...

See also: Full contents · Preview chapters · Illusions in the Boardroom