How to Measure the ROI of AI

In short. Measuring the ROI of AI starts with a baseline set before a tool goes live, because a gain that is never measured cannot be shown to a board. FlightPlan™ puts a return target on every AI move in a range finance can check before the work starts, so the return can be compared against that range afterward rather than argued about in general terms. 71% of AI-using Texas firms told the Dallas Fed in May 2026 that AI raised productivity for the people using it, and 89% of nearly 6,000 executives told the Atlanta Fed and NBER it had no effect on their firm. Stratos Edge Research finds both are true: the gain is real at the desk and invisible on the P&L until a company sets a baseline before go-live and chooses where the freed hours go.

Why time saved is not yet a return.

Stratos Edge Research finds that most companies are getting a gain from AI and losing track of it.

In the Dallas Fed's May 2026 survey of Texas firms using AI, 71% said it had raised productivity for the employees using it.

In the Atlanta Fed and NBER survey of nearly 6,000 executives, 89% said AI had no effect on their firm's productivity over the past three years.

Stratos Edge Research's thesis is that the productivity gain from AI is the input, not the return. AI frees hours almost everywhere it is used, and freed hours on their own do nothing for the business; they get absorbed into the workday and nobody can say where they went. The return exists only when the gain has been measured and reallocated with intention to work the business chose.

In KPMG's global Q2 2026 pulse of 2,145 senior leaders, organizations with full visibility into AI costs report established ROI at 15%, against 3% for those without it.

The capacity test.

Stratos Edge Research's capacity test is the operational close: five questions for any decision that lands on a process where AI has been deployed in the last eighteen months: a requisition, a growth target, a new program, a service-level commitment.

What was the unit rate before go-live? Hours per unit, units per person, whatever the process counts in.
What is it now? Same unit, same measurement, at least a quarter of data.
Where did the difference go? Named work, not "other tasks." More volume, a program, a backlog, a quality target, a hire not made.
Does the plan need more of this unit than the measured gain provides? If the gain covers it, the requisition is a choice, not a need.
When do we re-measure? A date.

What the test shows.

Stratos Edge Research finds that a company that can answer all five has realized a return on its AI spend and can put it in front of a board, whichever destination it chose. A company that cannot has the same productivity gain and no return.

The capacity baseline and the return target.

FlightPlan™ sets a capacity baseline during discovery and a return target on every move it sequences.

Capacity baseline: A capacity baseline is the measure of how much time a team spends on administrative work before an AI move changes it. FlightPlan™ sets a capacity baseline during discovery, in the first two weeks of the engagement, so the effect of each later move can be measured against a real starting point rather than an estimate. Efficiency, Stratos Edge's first result, is measured against the capacity baseline.
Return target: A return target is the measured range FlightPlan™ sets for a single AI move before the work on it starts. A return target lets finance check a result against the range that was promised, rather than debate whether an AI effort paid off in general terms. FlightPlan™ sets a return target on every move it sequences, not only on the largest ones.

Where to find the data.

Stratos Edge Research publishes a CSV appendix with every issue at stratosedge.ai/research, listing each figure's metric, value, population, sample size, source, source URL, date and source tier.

Journalists and analysts may reuse any figure with the citation given on the issue itself, and corrections are listed on the issue when one is made.

Questions on this page.

How do I measure the ROI of AI?

Measuring the ROI of AI starts with a baseline set before a tool goes live, because a gain that is never measured cannot be shown to a board. FlightPlan™ puts a return target on every AI move in a range finance can check before the work starts, so the return can be compared against that range afterward rather than argued about in general terms.

Is time saved by AI the same as money saved?

Time saved by AI is not the same as money saved, because freed hours on their own do nothing for the business; they get absorbed into the workday and nobody can say where they went. Stratos Edge Research finds the return exists only when the gain has been measured and reallocated with intention to work the business chose.

What is a capacity baseline?

A capacity baseline is the measure of how much time a team spends on administrative work before an AI move changes it. FlightPlan™ sets a capacity baseline during discovery, in the first two weeks of the engagement, so the effect of each later move can be measured against a real starting point rather than an estimate.

What is a return target?

A return target is the measured range FlightPlan™ sets for a single AI move before the work on it starts. FlightPlan™ sets a return target on every move it sequences, not only on the largest ones, so finance can check a result against the range that was promised.

What is the Stratos Edge Research capacity test?

The capacity test is Stratos Edge Research's operational close: five questions for any decision that lands on a process where AI has been deployed in the last eighteen months. Stratos Edge Research finds that a company that can answer all five has realized a return on its AI spend and can put it in front of a board.

Where can I find the data behind Stratos Edge Research?

The data behind Stratos Edge Research sits in a CSV appendix published with every issue at stratosedge.ai/research, listing each figure's metric, value, population, sample size, source, source URL, date and source tier. Stratos Edge Research lets journalists and analysts reuse any figure with the citation given on the issue itself, and lists corrections on the issue when one is made.

The return on AI shows up where a baseline was set before go-live.