Sreedhar Peddineni, CEO and Co-Founder of GTM Buddy.
Last month, I wrote that most enterprises are funding an AI transformation they cannot yet define. The most common response I heard from executives was a fair follow-up: “Fine. But how do I know whether ours is actually working?”
I have started companies through three major technology category shifts, and I have learned that when a leadership team cannot tell whether a transformation is working, the problem is rarely the data. It is usually the unit in which the data is measured. So, before you commission another dashboard, try an exercise that costs nothing and is often uncomfortable.
Put two documents side by side.
The first is your annual revenue plan, the one that justified your team to the board last fall. The second is the operating dashboard you looked at this morning.
The plan is written in capacity. You may not use that word, but look at the math: number of sellers, times ramp curve, times quota, times expected attainment, equals coverage. Whatever your operations team named the spreadsheet, it is a model of how much revenue the team is structurally able to produce, its revenue capacity. Every revenue leader I know builds one, every year, because no board approves headcount without it.
Now look at the dashboard. Calls logged. Meetings booked. Tool adoption. Hours saved by AI. Somewhere between the November board deck and the January operating review, the unit of measurement changed, from what the team can produce to how busy the team is, and nobody announced the switch.
That switch was survivable for a decade because activity and revenue moved in tandem. AI just broke the correlation, and that is why the contradiction suddenly has a price.
The productivity paradox is a measurement error.
Where AI is genuinely used in selling, it returns real time. Gartner’s May 2026 survey of 210 chief sales officers puts the figure at 4.8 hours per seller per week, close to a full workday, yet finds that 72% of sales organizations report low reinvestment of that reclaimed time into high-value activities. Gartner’s analysts call the resulting gap the sales productivity paradox: efficiency rises, results do not, because the organization absorbs the gains.
This is a unit error.
“Hours saved” is a substitution metric. It tells you a task got faster. It does not tell you whether a seller can now carry more revenue, and those are different questions with different answers. A time-savings number that climbs while output stays flat is not evidence of progress. It is a precise measurement of capacity being freed and then lost, taken with an instrument that cannot see the loss. Your dashboards are not lying to you. They are answering the wrong question fluently.
The academic evidence points the same way. Stanford’s Digital Economy Lab examined 51 enterprise AI deployments that had already moved past pilot stage into measurable value. Most of them, the researchers found, ended up being scored as cost reduction, while the largest returns clustered among the companies that aimed AI at revenue instead. The ambition is close to universal and the achievement is not: Deloitte’s survey, cited in the same report, finds roughly three-quarters of organizations hoping to grow revenue through AI and one in five actually doing it. Read that as a natural experiment in unit selection. The teams that measured what got cheaper got savings. The teams that measured what the business could produce got growth. The unit you choose is not just how you observe the transformation; it is part of what decides its outcome.
This is the definition test I promised last month, made concrete: an AI transformation you have actually defined is one measured in the same unit as your plan. If your plan speaks capacity and your operating reviews speak activity, you do not have a measurement problem. You have two languages and no translator, and the transformation budget is being spent in the gap between them.
Three moves to reconcile the units.
1. Make the plan’s unit the operating unit: The capacity model you build every fall should not be an annual ritual that gets filed until the next planning cycle. Keep it live. The question in the weekly review changes from “Were we busy?” to “Did what each seller can carry actually move, and are we realizing it?”
2. Ban “hours saved” as a terminal metric: Time saved is an input, not an outcome. Require every AI initiative to report one step further: where did the hours go? If the answer is “into more selling, and here is the evidence,” fund it. If the answer is silence, you have found where the transformation is leaking. Gartner’s own data suggests the difference is not marginal, organizations that reinvest reclaimed time are more than twice as likely to beat their customer growth goals as those that do not.
3. Track a ceiling and a gauge, not just a target: Every leader can recite quotas from memory. Very few can say what each seller could produce if the friction around their work were removed, or what share of that potential is being realized today. Those two numbers, kept current, are what make an AI investment defensible in a board meeting, because they are written in the board’s own language.
The question for your next operating review.
None of this requires a purchase, and all of it protects the spend, which is the standard I hold every recommendation to in this territory, because the vendors, mine included, have an interest in your confusion resolving in their favor. Resolve it in yours first.
Last month I asked whether you are funding an AI transformation you have defined or funding it because everyone else is. Here is the sharper version of that question, now that you can test it: is your transformation measured in the unit your plan is written in?
If it is, you will know whether it is working. If it is not, no dashboard, however green, will ever be able to tell you.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?