Pick the Number First: Measuring AI by Hours Saved and Work Won

“Everyone seems to like it” does not survive a budget review. Name the number before you build — hours per response, calls booked, days to close — and AI stops being a vibe and becomes a line item you can defend.

Most AI projects fail not because the technology doesn't work, but because nobody can say what it changed. “Everyone seems to like it” feels good in a hallway and disappears in a budget review. The fix is almost embarrassingly simple: pick the number before you build.

Name the number in advance

If you can't name the metric before you start, you won't be able to defend the spend later. The number doesn't have to be big — it has to be real, and it has to be visible to someone who signs things. A few that hold up:

• Hours per RFP response (prep time before vs. after).
• First calls booked per week per rep.
• Days to close the books at month-end.
• Time-to-first-draft on the work that bottlenecks everything downstream.

Hours saved or work won — not vibes

Good AI metrics come in two flavors: hours saved (the same work, faster) and work won (more output or better outcomes from the same team). Both are concrete. “Adoption,” “engagement,” and “sentiment” are not — they're proxies that tell you people are clicking, not that the business is better off.

Measure the baseline before you launch

You can't prove improvement without a starting point. Spend a week measuring the workflow as it runs today — how long it takes, how often it's done, where it stalls. That baseline is what turns “it feels faster” into “we cut prep time 60%.”

Keep the number small enough to be true

Resist the temptation to inflate. A defensible 20% beats a fantastical 10x every time, because the 20% survives scrutiny and the 10x invites it. The goal isn't a headline — it's a number a CFO can repeat without flinching.