Nine AI ideas, and one quarter's calendar that kills almost all of them. For a services business the sharper question is narrow: which two workflows can change revenue or margin inside that quarter?
The trap
A new VP of ops shows up with a list of every AI idea she has heard about. Intake agents, proposal drafting, call summaries, support bots, CRM cleanup, forecasting, knowledge search, content, reporting. Nine items, and they all sound useful. Then the quarter's calendar gets involved, and almost none of them survive it.
The work fragments. Each project picks up a demo, a pilot owner, and its own Slack channel. None of them ever gets operational enough to matter.
The scorecard
We score each idea on leverage, whether the data exists, who owns the workflow, how fast it reaches first real use, and how directly the output touches revenue or gross margin.
That usually leaves two projects standing. One removes a bottleneck that slows down sales or delivery. The other builds a repeatable operating loop the team keeps running after we leave the building.
The handoff
A shipped project needs four things: a named owner, a runbook, a failure path, and one weekly metric the owner actually checks. Miss those and you shipped a demo.
Five decent prototypes lose to two tools that operators open every Monday and trust.
