77 ideas, 7 worth building

A thirty-person creative agency in Australia. I measured how they already used AI, found where their hours actually went, and turned both into a ranked build plan priced on their own figures.

Twenty-eight of twenty-nine people already used AI most days, and nothing one person worked out ever reached anyone else. Four teams had each built their own way of tracking work. Nobody owned any of the sixty-six software subscriptions.

Stage	What I did	Output
Listen	Ran discovery sessions with the directors and all five teams.	Six sessions and a P&L call
Survey	Built and ran a staff survey on how each person uses AI today.	29 of 29 answered
Measure	Pulled a month of timesheets from their system and split billable from non-billable time by team.	3,624 entries
Audit	Reviewed every software subscription for use, overlap and ownership.	66 platforms, none with an owner
Rank	Scored every build candidate on one spine and priced the payback on their own cost rates.	77 found, 7 recommended

• The rollout starts where the rest of the company waits. Account Management held nearly half of all non-billable hours.
• The largest block of time, making the work itself, was left alone. The plan takes out the fetching, re-keying and chasing around it.
• Ten things they could start that week at no cost, beginning with a written rule on what client material may go into which AI tool.

The whole diagnostic, whether they carry on or not. The number, the evidence behind it, and the ranked plan. It was priced on their figures, so they can check every line of it themselves.