The automation audit that fits in one afternoon

Most automation engagements open with a "discovery phase" that takes three weeks and produces a slide deck nobody re-reads. The useful version fits in an afternoon: walk every process that touches a customer or a dollar, and time how long a human currently spends on each step.

The audit has one rule: if nobody can explain the current process without saying "it depends" more than twice, it's not ready for automation; it's ready for a decision. Automating an undecided process just makes the indecision run faster.

I score each candidate on two axes only: frequency (how often it happens per week) and variance (how often the steps actually change). High frequency, low variance wins every time: that's where an automation pays for itself before the invoice is due.

The output isn't a roadmap with thirty items. It's usually three: one that ships this week, one that needs a data source fixed first, and one that's honestly not worth automating yet. Clients trust the "not yet" more than any pitch deck.