A board member asks the question every CEO running an AI program eventually gets: "What's the ROI?" Most people answer with a story or a hand-wave, because the honest answer is that AI ROI rarely lands as one clean number. It shows up as a faster monthly close, a tighter forecast, fewer revisions on a deliverable, a deal that didn't get dropped. That doesn't mean it can't be measured — it means you need a better framework than the spreadsheet you used for the last software purchase.
Three buckets, not one number
Sort the value into three buckets and measure each on its own terms. Time saved: hours per week reclaimed on a specific task, multiplied by what those hours cost or what was done with them instead. Revenue moved: deals closed faster, proposals out the door sooner, more cycles through the sales funnel. Quality improved: fewer errors caught downstream, fewer rework loops, fewer client escalations. Mixing these into one figure is what turns ROI conversations into theater. Keeping them separate is what makes the conversation real.
Pick a baseline before you start, not after
This is the unglamorous step almost everyone skips. Before you roll out the tool, write down what the current state actually is: how long the monthly close takes today, how many revision rounds a proposal goes through, what the error rate on data entry is. Without a baseline, every post-rollout number is a guess dressed up as a measurement. With one, even a rough one, you have something to defend.
Be honest about timeframes
Software ROI is usually visible in a quarter. AI ROI often shows up over two or three. The first quarter is mostly people learning what the tool can do; the second is workflows changing around it; the third is when the real productivity shift gets visible in the operating metrics. If your board wants 30-day results, manage that expectation before you start, not after the first review.
The honest caveat
Not every AI investment will pencil out, and pretending otherwise damages credibility. Some pilots will be flat. The discipline isn't proving every project worked — it's being able to tell, faster than your competitors, which ones did and killing the ones that didn't.
Your next step
Before your next AI rollout, write a one-page baseline: the three metrics you'll watch, the current numbers, and the date you'll check back. That single page does more for ROI conversations than any vendor case study ever will.