Everyone already has a monitoring dashboard. Except the number just sits there,
red, for three months, and nobody knows which of the forty changes that quarter
pushed it up. It is almost never the code your team wrote. It is a font, a
third-party tag, an image that shipped without anyone measuring it.
With AI it is the same story. The agent passes the demo and falls over once real
people arrive, the API bill doubles with no explanation, and the
retry someone added to fix it doubles the bill again. The
start is always the same: measure before touching anything, in an environment that
resembles production, because optimising against lab numbers fixes the lab. After
that, the rest is work.
The calendar belongs to one person, so not every project fits. When it does not
fit, or when it is not the kind of problem I solve, I say so in the first
conversation. That is cheaper for both of us than finding out in week three.
Talk about a project ›