Then AI Arrived
Two insurers, same starting point. One uses AI to write a better board pack. The other uses AI to discover whether the board pack matches reality. Twelve months later, the gap is structural and compounding.
Every software-dependent corporate now faces a choice that sounds technical and is really structural: point AI at the narrative, or point it at the system.
Imagine two insurers starting in roughly the same place. Same market. Same regulation. Same legacy stack. Same executive habit of calling dysfunction “complexity.”
Insurer A chooses the writer path. The board pack gets better. Status reporting gets faster. The board feels more informed, which is not the same thing as being more informed.
Insurer B does the slower thing first. One process, one owner, one definition, one set of contracts, and only then the machine. The early output is not flattering: discrepancies, missing definitions, drift, awkward questions.
Twelve months later, Insurer A has a better board pack. Insurer B has a claims unit that resolves claims materially faster, at lower cost per claim, deploying continuously instead of quarterly. The weekly synthesis catches discrepancies and resolves them within the week. The engineering team’s attrition has dropped to near zero.
The most important difference: Insurer B can know its own numbers without assembling a heroic interpretation layer. The artefacts connect investment to outcome. Insurer A still cannot produce its own equivalent picture because the measurement would require tracing a process nobody has written down through a system nobody owns.
The fork is a governance choice.
The writer path deepens dependence on the problem it fails to solve. Each quarter creates more narrative infrastructure: more AI-generated updates, more polished summaries, more roles whose status depends on the narrative remaining the thing that counts as evidence.
Moving from the writer path to the reader path is not a tooling migration. It is a withdrawal of confidence from the machinery the organisation has spent years building around presentation. Delay is not neutral. Delay is construction.
On the reader path, the opposite dynamic holds. Each reconciliation makes the next one cheaper. Definitions improve. Contracts sharpen. The organisation’s self-description converges on its system rather than away from it.
| Writer | Reader | |
|---|---|---|
| AI generates board pack content | • | |
| AI compares strategy against code | • | |
| Process definitions are prose or slides | • | |
| Process definitions are structured | • | |
| Units interact through meetings | • | |
| Units interact through contracts | • | |
| Discrepancies found in quarterly reviews | • | |
| Discrepancies found in weekly synthesis | • | |
| Narrative improves each quarter | • | |
| Operations improve each quarter | • | |
| Best engineers leave | • | |
| Best engineers stay | • |
Count the marks in each column. If you have more in the Writer column, AI is accelerating your fiction. If more in the Reader column, AI is accelerating your operations. Most organisations will have a mix, and the mix tells you where to start.
The next article, The Process Is the Product, is what the winning side of that fork looks like in practice. Where This Is Hard is where the counterarguments are strongest.
See also: All articles · Illusions in the Boardroom · Illusions of Work