Structural Correction

Chapter 14: The Transitional Authority

The structural investment required cannot be funded through the existing budget model. A dedicated transitional fund under the CTO’s authority, with a hard twelve-month cap, resolves the bootstrapping deadlock while preventing concentrated authority from becoming permanent.

A European retail bank authorised structural change but routed boundary decisions through a cross-functional steering committee rather than a CTO-CEO decision pair. The committee comprised representatives from technology, operations, risk, finance, and HR. Each had legitimate concerns. Each had veto authority in practice if not in charter.

Three months in, the first autonomous unit launched: payments, clean boundaries, quick results. Six months in, the second unit's boundary decision required splitting a shared customer database. The steering committee requested impact assessments from four affected teams. Each team submitted concerns. The committee scheduled a follow-up. The follow-up was delayed because two committee members were on leave. By month nine, the boundary decision was still pending. The unit operated without clear ownership of its data, reading from a shared store it could not modify, publishing through interfaces it could not version.

By month eleven, the engineer who had been carrying the customer data model left. She had held the mapping between the old schema and the new unit's needs in her head. The unit launched without her knowledge. It failed its first regulatory audit because the data lineage could not be traced through the split that never happened. The board noted the failure and commissioned a review. The review recommended a programme. The audit committee at the insurer described in the previous chapter took a different approach.

The audit committee reconvened six weeks after Elena's first set of questions. In the intervening weeks, the CEO had called her privately to say the CTO felt ambushed. Elena had expected this. She had not expected the fellow NED who told her, over coffee after a charity event, that the committee was “not the venue for that kind of interrogation.” She noted it and moved on. She had received two documents in advance.

The first was the structured process definition for claims management: states, transitions, decision points, failure modes, version 0.1. The CTO had produced it in three weeks with a principal engineer and two domain experts who had been carrying the process in their heads for years. The document was imperfect and acknowledged as such, forty-three pages with seventeen items flagged as “requires validation against production behaviour.” It was also the first description of the claims process that could be tested against the code, and the reconciliation had already surfaced four places where the documented behaviour and the actual behaviour diverged.

The second document was a budget request. The CTO had estimated the cost of establishing the first four autonomous units: decomposing shared services, establishing contract boundaries, staffing units, building the observability layer that would produce the reconciliation evidence Elena had demanded. The figure was € 6.2 million over twelve months, roughly three times the normal annual central technology budget for structural work.

...

See also: Full contents · Preview chapters · Illusions of Work