Margin erodes quietly. Repeated low-value tasks, rework loops, and review gaps each take a small bite at a different project stage, and the practice feels the total at the end as a number, long after the causes have passed. Diagnosing those causes from memory tends to find the loudest irritation, which is often not the most expensive one.
A short structured diagnostic for a practice owner. Choose the project type and stage, rate the time drains and rework causes, and it returns a margin-leak map: where the time is going, what each leak looks like, and which kind of intervention fits it.
The whole pass takes under five minutes. The output separates AI opportunities from ordinary process fixes, because plenty of leaks need a template or a checklist rather than a model, and it ends with one recommended first experiment instead of a long list. The result exports as a short action note the owner can take to a practice meeting.
The diagnostic works from what the owner tells it, and it says so. The leak map is a structured reading of self-reported pain, useful for choosing where to look first, and the suggested fixes are starting points for a person to test against real timesheets and project records before anything changes.
Whether the margin conversation can start from structured evidence instead of anecdote, and whether a five-minute diagnostic is enough to give a practice owner a defensible first experiment rather than a vague intention to "use more AI".