Structural Correction

Chapter 17: What the Sceptical Board Member Asks

Five adversarial objections to the transitional fund renewal, each individually rational, each answered with evidence rather than argument. The chapter also honestly names where the model has genuine limits.

At the quarterly review where the transitional fund renewal was on the agenda, the claims unit had three quarters of data and the renewals unit two. A non-executive director named Thomas, quiet through the previous reviews, opened his folder. He had spent thirty years in risk and underwriting and seen five transformation programmes arrive with confidence and depart with lessons learned. He had five concerns, and he intended to raise them before the fund was renewed.

“Before we discuss the renewal,” Thomas said, “I want to understand the technology dependency. You are asking us to keep reorganising around a capability that is still maturing. What happens when the AI gets it wrong?”

Elena did not argue; she placed the reconciliation report on the table. “The question is not whether the AI is ready but whether we can describe a single critical process in machine-readable terms. Can we?”

The CTO answered that they could: the claims unit's process definition was in its fourth version, and the reconciliation had surfaced four divergences in the first cycle that no human audit had found, a timeout counting calendar days instead of business days, an undocumented re-review step used forty times a month, a silent exemption for low-value claims, and a routing rule that contradicted the published escalation policy. The structural work was the precondition; the AI was the instrument. The sharper question was whether it was acceptable, regardless of AI, to operate critical processes that no one could describe, that existed only in the heads of people who might leave, and whose actual behaviour had never been compared to their documented behaviour.

Thomas paused. He did not concede, but he moved on. “I have been through four restructurings. Each promised clearer accountability. Each produced new names for the same problems. Why is this different?”

Elena asked the claims unit lead to present three things. First, his calendar: before the transition, forty-one hours of meetings a week coordinating across six teams; after, eleven, most internal to the unit. Second, a contract dispute with the payments unit resolved in three days through the five-day resolution clause, which under the previous structure would have taken a steering committee and six weeks. Third, the unit's deployment cadence: forty-three deployments in the first quarter, each without cross-team coordination. “The test,” the unit lead said, “is whether a unit can deploy a change to its process without coordinating with another team. If yes, ownership is real. If no, the reorganisation is cosmetic. We can.”

...

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