112% Higher AI ROI — But Only 25% of Tech Executives Measure It Consistently

IBM recently published two numbers that should make almost every executive managing AI investment stop and think.

Organizations that use AI performance data to inform funding and talent decisions report 112% higher AI ROI than their peers. Yet only 25% of tech executives consistently measure AI ROI.

That is a remarkable gap. The obvious lesson is that measurement matters. But I think the more interesting question is:

What if the biggest value of AI ROI measurement isn't knowing what happened—but knowing what to fund next?

ROI is not a reporting problem

It is easy to treat ROI as something that gets calculated after the work is done. A project launches. The team reports results. Finance reviews the numbers. Someone decides whether it was successful.

But IBM's findings point toward a much more useful role for measurement. Performance data should influence funding and talent decisions. That means measurement is not just a scorecard. It is part of the operating system.

The sequence starts to look like this:

Define value → measure performance → evaluate evidence → make a decision → reallocate resources → measure again

That last step matters. Because the process should repeat.

The real payoff is the next decision

IBM recommends connecting AI initiatives to specific business outcomes before significant resources are committed. Leaders should define how AI is expected to affect the outcome, establish how progress will be tracked, generate evidence early, and then decide whether continued funding is justified.

Some initiatives should scale. Others should be adjusted. Some should stop altogether. IBM's stated goal is simple: evidence should determine where AI investment goes next.

That is a much more powerful use of ROI than simply proving that a project worked. A good measurement system helps answer:

• What deserves more funding?
• What needs redesign?
• What should be constrained?
• What should stop?
• Where should talent move?
• Which assumptions are no longer holding?

The value of measurement is not the number itself. It is the decision the number makes possible.

The cost of not measuring is not just uncertainty

IBM reports another striking number: 68% of Chief AI Officers say they launch AI initiatives even when they have no reliable way to assess or measure impact.

That creates a deeper problem than weak reporting. If an organization cannot reliably assess impact, it also has a weaker basis for deciding what to do next:

• A weak initiative may continue receiving resources because nobody can demonstrate that it is underperforming.
• A strong initiative may remain underfunded because its value is not visible.
• Talent can stay attached to low-value work.
• Capital can be spread across too many projects.
• Weak assumptions can survive simply because they were never tested.

The problem compounds at the portfolio level.

AI will make this cycle move faster

This is where the issue becomes more urgent. AI is increasing the velocity of experimentation, deployment, output, and change. Business cases, model costs, capabilities, workflows, and new use cases will all shift faster.

That means the decisions around AI investment will also need to happen more frequently:

• A portfolio that was reviewed quarterly may need more continuous adjustment.
• A model that made economic sense six months ago may no longer be the right choice.
• A workflow that performed well last quarter may now have a cheaper alternative.
• A successful pilot may deserve capital sooner.
• A weak initiative may need to be stopped before it consumes another budget cycle.

The feedback loop gets tighter:

measure → decide → reallocate → learn → measure again

As AI speeds up the business, slow measurement becomes more expensive.

Measurement creates learning velocity

There is another reason this matters. Every AI initiative produces information about what works. The question is whether the organization captures that information and changes its behavior because of it.

• Did the expected value appear?
• Were the assumptions right?
• Did the workflow improve?
• Did costs rise?
• Did quality hold?
• Did the investment deserve to scale?

If the answers do not affect the next funding decision, the organization has data, but it has not necessarily learned. Learning happens when evidence changes action.

That is why ROI measurement should not sit at the end of the process. It should sit inside the decision loop.

The 112% gap may be telling us something bigger

IBM is careful about causality: its report synthesizes results from multiple studies with different samples and designs, so the findings are intended to show directional patterns rather than prove that measurement alone causes higher returns. That caveat matters.

But the pattern is still hard to ignore. Organizations using AI performance data to guide funding and talent report dramatically stronger returns. Most tech executives still do not measure consistently, and a large share of AI leaders launch initiatives without a reliable way to assess impact.

That suggests the AI ROI problem may not be only about technology. It may also be about whether organizations have built a disciplined way to turn evidence into better resource decisions.

Because the point of measuring AI ROI is not to produce a number. It is to improve the next decision.