Reference

The Canon

Numbered patterns, grouped by the problem each one answers rather than by the date it was written. Every entry opens with the summary a chief executive can act on and closes with what to do about it.

8 entries
Before anything is built

I have the mandate and nowhere to point it

  1. 006

    A mandate is not a question

    Do something with AI is an objective, not a problem statement. An objective cannot be solved, only performed.

    Read entry →
After the pilot works

The pilots work. Nothing scales.

  1. 001

    Pilot purgatory is a funding artefact

    The block is the budget, not the technology. Pilot money and production money come out of different pockets.

    Read entry →
  2. 004

    Permissions set the ceiling, not model quality

    Most complaints that the model is not good enough are complaints that the assistant cannot reach the document holding the answer.

    Read entry →
When finance asks

I cannot show what any of it paid for

  1. 002

    The denominator problem in AI return figures

    Return figures are not comparable between organisations, because almost nobody publishes what went under the line.

    Read entry →
  2. 005

    Headcount-neutral is the only honest first-year claim

    Promising role reduction in year one buys a commitment finance will hold you to and a threat staff will organise against.

    Read entry →
When the tool is live and quiet

My people are politely refusing

  1. 003

    Shadow adoption runs ahead of sanctioned adoption

    Your baseline is not zero. Staff were using consumer AI for work long before procurement started.

    Read entry →
  2. 007

    Refusal is usually a measurement signal

    Refusal is arithmetic, not attitude. If checking the output costs more than the task saved, stopping is the rational move and encouragement will not change it.

    Read entry →
When the board asks

I am being asked what could go wrong

  1. 008

    What boards actually ask

    Boards are not evaluating whether the technology works. They assume it does. They are evaluating whether the organisation is exposed.

    Read entry →