The AI ROI Question Nobody Asks: Where Do the Gains Go?
Your team is already faster with AI. The question almost nobody asks is where those productivity gains are going. If you don't direct them, your employees will decide for you.
In short
AI is making individuals and teams faster, but few companies know where the saved time is going.
- If you aren't explicit, employees will decide for themselves how to use that capacity.
- The right destination depends on what your company needs most.
Where are your AI gains going?
One of the most common questions I get from friends and colleagues right now is some version of, "What does our AI ROI look like?" I wonder if they shouldn't be asking a different question: "Where are our AI productivity gains going?"
Most of the people asking run small and mid-sized businesses that existed well before AI did and are now somewhere in the middle of an AI transformation. They've bought the tools. They've watched their teams get faster. What almost none of them have done is decide what that faster team is supposed to do with the time.
At Study.com (sister company to Candova), we've been on this journey from the beginning, and given how fast technology moves, I don't expect it to end. If you can't point to any benefit at all yet, that's an implementation or measurement problem and a different article. My colleagues have written on Candova.ai about what to track instead of seat licenses and what AI actually costs once you include training time and the leadership hours to redesign how work gets done. Start there. This piece assumes the gains are real, and discusses how you're using those gains.
Three employees, one productivity gain
Imagine three curriculum managers who create courses and lessons for a living. I'll use curriculum because it's a common use case we see day-to-day, but we could create similar examples for marketing, engineering, design, finance, or the people team. These managers have the same role, tools, and job description, and nobody has told them what to do with the time AI just gave back.
One curriculum manager says: "I used to create five new lessons a week. Now I can create ten."
A second curriculum manager says: "I still create five lessons a week, but they're more detailed and higher quality than what I used to produce."
A third curriculum manager says: "I used to spend all my time creating lessons. Now I create them in half the time and spend the rest exploring how AI can help students when they get stuck on practice exercises."
Which one created more value? All three can now create five lessons more quickly. What they do with the remaining time is the interesting question.
If you're capacity-constrained, the first employee's approach is worth the most. If you're losing deals on quality or paying for rework, the second one is. If you're in a fast-paced, innovation-driven market, the third is probably best. Maybe you need some combination of all three. The right answer depends on what your company needs now. All three made a reasonable choice. All three made it alone.
The four places AI productivity gains can go
If you are not explicit about where your productivity gains are going, your employees will decide for you.
At any given time, you have a strategic decision to make about where these gains are directed, and there are four places to put them:
- 1More output. Same work, more of it.
- 2Better output. Same volume, higher quality.
- 3New work. Capacity redirected to things you couldn't staff before. Innovation lives here, which is what our third curriculum manager chose.
- 4Nowhere. Time is lost to the inertia of the day through more meetings, more documentation, and other work that wasn't deliberately prioritized. (Yikes!)
And people are not refusing to be directed. They're waiting to be. In our own research across 1,000 employees, only 32% said they have a clear standard for what good AI use looks like in their role. The other 68% have a rough idea at best, and one in five said they don't know at all.
By now, you may be screaming that I missed the most obvious option for AI productivity gains: cost savings! The list above assumes we're retaining capacity within the organization. Reducing headcount or avoiding future hiring is a separate corporate allocation decision; this framework addresses where to direct retained capacity.
Be explicit, by team and by role
We try to specify where the gains should go, by team and by role, before the gains arrive. Honestly, we still have work to do to determine how this type of expectation-setting aligns with our normal business objectives and desired outcomes.
For product development (e.g., engineering, design, product management), we want to deliver more value to users more quickly, given our current resources. This means that, for a given team and time period, we want to solve more customer problems for users while still meeting our base-level quality requirements for releasing new features. One internal diagnostic we use is story-point analysis, alongside manual reviews of our roadmap and Jira data to assess the problems solved and value delivered.
For customer service, we want to use AI to improve customer interactions with a fixed set of resources. We can evaluate performance through response speed, personalization, first-contact resolution, time to final resolution, and other measures. We have found that customer service satisfaction is a key component of users' brand perception and overall product satisfaction.
Within our content team, we want to create time for innovation. Other than engineering, no group in our company is more affected by the potential of AI. That means our content team needs the time and space to experiment with new content formats and explore how AI can improve learning outcomes for our users.
Next steps
Given that, where should you start?
- 1Pick a team where you're confident you're seeing productivity gains but are unsure how people are using the extra time.
- 2Decide with the team where those productivity gains should go.
- 3Choose a metric or measurement process that shows where the gains are going.
Common questions
How do we know if the gains are going nowhere?
Look for the absence of a change you'd expect. On our product teams, comparable features should require fewer story points than they used to. If that measure is flat a couple of quarters into an AI transformation, the time went somewhere else. The harder case comes later, once the easy speedup is behind you and story points stop telling you much. Then you're asking a different question: are we solving bigger problems than we used to? If the answer is no, and nothing else changed either, the gains were absorbed.
Isn't this the same as measuring AI ROI?
They're related, but not the same. Measuring ROI tells you what happened. This is about deciding what you want to happen and, therefore, what you will be measuring in your ROI analysis.
Is this different for small and mid-sized companies?
The decision is the same, but it matters more. Large organizations can fund dedicated measurement infrastructure to spot misallocated capacity after the fact. Smaller companies usually can't, which makes naming the destination up front the highest-value move available. It's the one that costs nothing but leadership attention.
Find out what your saved hours are worth
Put a number on the capacity AI is already returning to your team, then decide where it should go.
Power users save 10+ hours a week. Learn how.
The practical AI habits behind it, one a week.
Written by
Michael Schmier
COO & President of Candova
Michael has spent roughly three decades leading operations and product across consumer, enterprise, and education. He helped pioneer the virtual reality market at Samsung, led the content business at BabyCenter, and held leadership roles at startups in data analytics and sports technology. The through-line is execution: taking a strategy and making a whole organization run on it.