n.V7-2.01| Decentralize Until It Costs You
01| Decentralization is not a value we hold. It is an instrument we use, and like any instrument it has a range where it works and a range where it does not. We decentralize because decisions made close to the work are faster, better informed and cheaper to make than decisions made three levels away from it. We stop decentralizing at the point where speed or reliability begins to suffer, and we accept less decentralization when that point arrives. There is no ideological floor below which we refuse to go.
02| The reason is plain. IkoCiti exists to create impact in the barrio. Every structural question resolves against that one test: whichever arrangement produces more verified output in the barrio wins, regardless of how it reads on an org chart or how well it scores against somebody's governance manifesto. A beautifully distributed organization that ships nothing has failed the only test that counts.
03| This puts us out of step with both of the fashionable answers. One camp says the answer is flatness — no managers, consensus, the crowd decides. That works until something has to be decided on Tuesday. The other camp says the answer is a clean hierarchy — clear lines, clear authority, someone accountable at every level. That works until the person with the authority is four thousand kilometres from the barrio and has never met anyone in it. We are not looking for the philosophically pure structure. We are looking for the one that gets projects built.
04| The permanent structure is deliberately small. Ten divisions. Inside each division, a handful of groups — permanent capability teams, roughly seven per division, about sixty-five in total. Below the groups, operators running the day-to-day. That is the whole of IkoCiti's permanent organization. Divisions, groups, operators. Nothing else is permanent, and we intend to keep it that way, because permanent structure is the thing organizations accumulate without noticing and cannot shed afterwards.
05| Leadership runs on three levels at Global: the Board, the Global Management Team, and the Division Leaders. The detail worth pausing on is that the Global Management Team is not a layer sitting above the divisions. It is the tenth division — all ten division leaders sitting together, wearing the same hats they wear in their own divisions. There is no separate corporate centre with its own staff, its own budget and its own reasons to grow. We did that on purpose. A head office with employees of its own will eventually develop interests of its own, and those interests are never smaller next year than they were this year.
06| Everything that crosses a boundary is temporary. When work needs people from more than one group, we form a Circle, sized to the job: work inside a single group needs no circle at all; work spanning groups within one division forms a Division Circle; work crossing divisions forms an Iko Circle; work where Global people and local CITI people build something together forms a CITI Circle. A Circle forms, takes ownership of the project, and dissolves when the project is done or judged unsalvageable. It borrows its people from the divisions and gives them back.
07| The first of those four settings is the one that matters most, and it is the one organizations forget. Most work needs no circle. Coordination is not free — it is a tax paid in meetings, handoffs and waiting. Having discovered cross-functional teams, a great many organizations proceeded to apply them to everything, and produced a workforce that spends its days coordinating work it no longer has time to do. A single group with a clear job should be left alone to do it. We form a Circle when the work genuinely crosses a boundary, and not one boundary earlier.
08| The Circle carries real authority for the life of the project — it has call on the staff contributed to it, and the divisions that lent them do not outrank it. This inversion is deliberate. In a decentralized organization nobody owns a cross-division outcome, which means a cross-division project fails quietly and nobody's name is on it. Giving the Circle priority over its feeder divisions is how we put a name on it.
09| Distributed authority needs a return path, or it is just delegation with better branding. Ours is the escalation right: a 60% majority of qualified Ikosi can lift any major decision by the Global MT, a Division MT or a CITI MT to the Board for a final ruling, and decisions inside a Circle escalate the same way to the Division MT. Management never holds the last word. The point is not that the right will be used often — we expect it to be used rarely. The point is that everyone decides in the knowledge that it exists.
10| What we do not know is where the line sits. Nobody knows in advance how much decentralization a given organization can carry before reliability starts to slip, and we will not pretend we have calculated it. We will find the line by crossing it — a decision that takes three weeks when the barrio needed it in three days, a standard that drifts in four CITIs at once before anyone notices. The structure described in this book part is our best current answer, and we will move it when the barrio tells us it is wrong.
11| That is the rule, then, and the rest of this part is its application: decentralize until it costs you, and let the barrio decide when it has.
THE END
(word count: approx. 880)
