There has been a real craze over the last couple of years to use AI in every situation we possibly can. Can we use it in outbound? Can we use it to handle inbound support requests? Can we use it for forecasting, for coaching, for research? The answer to most of those is yes.
What we haven’t focused on nearly as much is establishing the return. And when nobody defined what success looked like before a project started, it’s very hard to prove it worked afterwards, which is usually when budgets get questioned.
Here’s the approach I use with the RevOps teams I work with.
Step 1: Pick the metric before you pick the tool
Before you start any AI project, decide which number it’s meant to move. A few common ones:
- Win rate
- Number of customer support tickets
- Conversion from lead to opportunity
- Revenue per headcount
- Average sales price
One project, one main metric. If you can’t name it, that’s a sign the project is a solution looking for a problem.
Step 2: Get an honest baseline
Measure where that metric sits today, before anything changes. It sounds obvious, but it’s the step most often skipped in the rush to get something live, and without it every result afterwards is a matter of opinion.
Step 3: Make sure the tool actually lines up with the metric
This is where the thinking really happens. Take revenue per head as an example. How is AI going to move it? One route is reducing the admin time it takes a rep to handle a single opportunity. If they spend less time on admin, they can handle more opportunities, and revenue per head goes up.
| If your metric is | AI should be changing | You’ll know it worked when |
|---|---|---|
| Revenue per headcount | Admin time per opportunity | Reps handle more opportunities each |
| Win rate | How reps run and follow up on deals | Win rates rise over time |
| Average sales price | How deals are positioned and negotiated | Average price trends upward |
| Lead to opportunity conversion | How leads are qualified | More leads qualify into opportunities |
| Support tickets | How inbound requests are handled | Ticket volume comes down |
Step 4: Compare the people who use it well
Adoption is never even. Some reps will use the tools far more effectively than others, and that difference is useful. Check whether win rates are climbing for the people using the tools well compared with everyone else, and whether those gains hold over time. If they are, you have your ROI story. If they aren’t, you’ve learned something important before spending more.
Consistent outputs make measurement possible
One thing that quietly undermines all of this is AI output that changes every time you run it. You can’t measure impact on a moving target. That’s why the groundwork of giving AI proper business context matters so much: without shared definitions for your fields, stages, and metrics, AI gives a different answer every time someone asks.
See it in action
Airspeed takes admin off your reps’ plates and shows you how deals, win rates, and team performance move over time, so you can tie AI to the numbers that matter. Book a demo to see it on your own pipeline.
Frequently asked questions
How do you measure the ROI of AI in sales?
Choose one business metric the project should improve, measure its baseline before you start, make sure the tool directly influences that metric, and track the change over time.
Which metrics show whether sales AI is working?
Common ones are win rate, revenue per headcount, average sales price, lead to opportunity conversion, and support ticket volume.
What is revenue per headcount?
Total revenue divided by the number of people on the team. AI can lift it by cutting the admin time each rep spends per opportunity, so they can handle more.
Why do so many AI projects fail to show ROI?
Usually because no metric or baseline was set before launch, so there is nothing reliable to compare the results against.
