Writing

No one has ever read every decision a company has taken, and a kept record with an AI reading it would be the first

Organisations remember what they decided and lose why within a few years, and the people who could reconstruct it have moved on by the time it matters. A kept record of the reasoning, read across functions and years, is new, because no person has ever held all of it at once. The valuable part is the record; the judgment on it is still yours.

AI tilts capital toward what can be measured, and the long bet loses on the business case

AI makes measurable improvements more visible, more frequent and better documented, so in any funding contest the ten-year decision now faces a stronger opponent than it did a year ago. Nothing in the model prefers the short term. The drift comes from the asymmetry in what can be shown, and the counterweight is a different kind of scrutiny with its own clock, not better judgment from leaders.

Every organisation has a resting cadence, and programmes do not change it

Organisations run at a speed that is produced, all the time, by how they are put together, and they go back to it once the extra pressure is gone. That speed changes for good when ownership changes, and hardly anywhere else.

The highest use of operating a technology yourself is the no

At board level, secondhand understanding of a fast-moving technology is a governance gap rather than a sensible delegation. The judgement that hands-on use buys is mostly negative: seeing early which technologies will not scale or do not fit the problem. It only arrives when the technology and the industry are understood at the same time.

Silence in a boardroom is a selection and chair problem

A board meets a handful of times a year and has nothing to build a culture in, so the work on psychological safety in teams does not carry over. Whether dissent gets voiced depends on who was selected and on what the chair charges for it at the point of decision; safety is what you measure afterwards, never the thing you do.

An AI that reports nothing found cannot tell you whether it looked

An AI system that reports "no exceptions" writes that sentence with the same confidence whether it searched the whole record or a fraction of it, and the output carries no trace of which. A constraint on the shape of the output holds that fault in check; a rule asking the model to judge its own work does not, and the facts it fails to join need a second pass.