Published on
Imagine a marketing team adopting an AI-enabled video platform.
The team can now produce more videos in less time. It launches additional campaigns and creates more content than it could before.
The productivity improvement is real. The people doing the work probably feel the difference.
Then the team presents the result to the business:
We produced twice as many videos.
That is useful information.
It is not yet proof of value.
Did the company close more business? Did engagement improve? Did the campaigns influence the result they were created to support?
This is the measurement problem organizations keep running into with AI. They track what AI is doing and assume that activity equals return.
It does not.
Measuring the return on AI investment requires looking beyond AI activity to what actually improved. Usage, output volume and productivity can show that people are using AI, but business value comes from better outcomes: stronger decisions, improved workflows, reduced risk, increased revenue or another measurable result tied to the organization's objective.
I call it ROAI
The question is return on AI investment.
What became better because the company used AI?
That sounds obvious, but companies are easily distracted by the metrics available inside the tool.
Token consumption can show adoption and cost. Logins can show whether employees are trying the platform. The number of outputs can show how much the system is producing.
None of those numbers proves that the business improved.
AI can produce a lot of garbage very quickly.
If output volume becomes the success metric, employees can hit the target while creating more noise for customers, coworkers and managers.
Measure the outcome, not the artifact
The marketing team may produce more videos.
That is an artifact.
The business outcome could be stronger engagement, better-qualified leads or additional closed deals.
The same distinction applies across the company.
A sales team may generate more account summaries, but the meaningful question is whether those summaries help representatives focus on the right opportunities.
A leader may receive more reports, but the question is whether the reports improve a decision.
A security team may produce more findings, but the question is whether risk is actually reduced.
The metric should follow the objective.
AI may create value by helping you choose
Suppose the marketing team uses AI to create three different campaign approaches and run them concurrently. AI then helps aggregate the performance data and identify the signals coming back from each campaign.
The team can use that information to decide where the company should invest more money.
That is not simply faster content production. It is intelligence-driven decision-making.
The business can track the decision, the investment and the result.
AI may create more value by helping an organization choose correctly than by helping it produce more.
Establish the baseline before the AI pilot
A lot of companies try to prove value after the experiment has already started.
At that point, they know what the AI tool is doing, but they may not know how the workflow performed before it was introduced.
Without a baseline, it becomes difficult to separate improvement from general activity.
Before changing the workflow, define how it operates today.
How much time does it take? What does the current result look like? Where does rework happen? What metric is the business trying to improve?
Then introduce AI and measure again.
This is basic implementation discipline. AI does not make it optional.
Measure AI training by what changes at work
The same principle applies to workforce education.
Completing AI training is an activity.
Using the capability to improve the work is the outcome.
The organization needs managers and business leaders involved because they understand what performance should look like after the learning is applied.
Learning and development can report that employees completed the program.
The business must determine whether the program changed anything that matters.
ROAI is not about finding one universal AI metric.
It is about refusing to confuse visible activity with business value.
Recent posts by Michael Burch
Stop Measuring AI Activity: Start Measuring ROAI
Imagine a marketing team adopting an AI-enabled video platform. The team can now ...
Managers Are the Missing Link in Enterprise AI Adoption
To find out whether AI training worked, do not begin with the learning platform. The ...
Redesign the Workflow, Not the Task
Every Monday morning, I get an email from myself. AI has reviewed my upcoming ...
