The Fix

Chapter 17: The Authority to Decompose

Scaling from one autonomous unit to many requires a transitional fund ring-fenced from the annual budget cycle and a CTO-CEO decision pair that collapses boundary decisions to two people. The chapter specifies how the CTO is measured and what remains central after the transition.

The CTO of a Nordic insurer had the first autonomous unit running for four months, with claims resolution time down 31% and deployment frequency tripled. The evidence was unambiguous, and she needed three more units by end of year to reach the tipping point she was counting on: the point at which the contrast becomes too plain to keep filing under “too early to tell,” and the question “should we restructure?” gives way, at least among the people who decide, to “why does everything else still work the old way?”

She opened the budget request for the payments unit. The decomposition work required € 1.4 million over six months: untangling the shared database that five product lines wrote to directly, establishing contract boundaries with three adjacent services, staffing the unit with a domain expert and two engineers who understood the payments regulatory landscape. The CFO's question was reasonable: which business unit's P&L should absorb the cost? Payments was shared across four product lines. No single product owner would fund decomposition that benefited the other three. The investment that would make the payments unit possible did not fit inside any existing budget line, because the budget model priced departments, not processes, and the payments process did not belong to any department.

So she escalated. The CEO agreed the investment was necessary and the CFO agreed in principle, but the process to secure cross-business-unit funding took eleven weeks, and by the time the budget was approved the senior engineer she had identified to lead the payments unit had accepted an offer elsewhere, carrying with him twelve years of payments domain knowledge that the organisation had never asked him to write down. The window for the second unit moved from Q2 to Q4. The claims unit continued to operate, producing evidence that made the surrounding dysfunction visible, while the surrounding dysfunction continued to operate, producing evidence that it could outlast any change that required its own budget model to fund it. Launching one autonomous unit requires selecting the process, drafting the governance charter, and protecting the unit's autonomy through the first ninety days. Extending from the first unit to the next, while the evidence it produces is still fresh, requires a budget and authority structure the existing model does not provide.

In some organisations the first act of decomposition is not funding a unit but stopping a programme that should never have been funded. The write-off is painful because it makes capitalised ambiguity visible; continuing is worse, because it converts missing business definition into more software, more integration, more local workarounds, and a larger write-off later.

The funding conversation changes when the CTO arrives not with a budget request routed through the normal cycle but with a reconstruction fund, pre-authorised by the CEO and sized to the decomposition work required. The CFO's first objection is structural rather than financial: which P&L absorbs a cost that benefits four product lines and that no product owner will fund on behalf of the other three? The fund resolves precisely that. It sits under the CTO's authority for a bounded period, ring-fenced from the annual cycle, with a twelve-month cap and a decline curve that transfers operational cost to the units as they become self-sustaining. The CFO's reporting obligation is met by a line, a cap, a taper schedule, and measurable outcomes attached to each draw; what the fund lacks is a product owner, because structural investment that requires one to justify it never gets justified, which is why the payments unit does not yet exist. If twelve months produce insufficient working units, the fund closes, and that constraint is the point, because an open-ended fund becomes a permanent dependency, whereas a capped fund forces sequencing discipline: the hardest decomposition first, the boundaries that unlock the most subsequent units, the work that cannot wait.

The payments unit began decomposition the following Monday.

The autonomous unit model requires modularity, and modularity requires structural investment: decomposing shared databases, establishing contract boundaries, untangling services that span multiple business processes, staffing units with the right combination of domain expertise and engineering capability. The investment is cross-cutting, its benefits accruing to multiple future units rather than to any single business function.

The existing budget model prices departments, not processes: it is an annual fiction in which money is allocated to names rather than behaviour, funding attached to products that cannot be located in the system. The contradiction produces a deadlock: the autonomous unit model needs modular systems, modular systems need structural investment, and structural investment needs a budget mechanism that does not exist inside the current model, so each element waits for the other two.

...

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