Every few months a prospect asks us to help them write an AI strategy. We say no, and then we offer them the thing they actually wanted, which is a list of what to automate first and what each one is worth.
Strategy is a stalling pattern
It is rarely deliberate. A strategy phase feels responsible, it involves everybody, and it produces a document nobody has to act on for a quarter. Meanwhile three people are still chasing invoices by hand.
What replaces it
Ten working days inside the operation. We shadow the work, map every workflow, and time what people actually do rather than what the process document says they do.
What comes back is a build plan with a price per agent and an hours figure beside each one. You can start at the top, or not start at all, and both are informed decisions.
The uncomfortable finding
A third of what we find is not worth automating. We say so in the handover, in writing, in front of the person paying for the audit. It is the fastest way to be trusted on the other two thirds.
Two exceptions
Below about thirty people the arithmetic rarely works, and a regulated process with no written rules needs the rules before it needs an agent. In both cases we say so on the first call.
002
/ Keep reading
From the journal