AI Strategy Without ROI Measurement: The Gap No One Owns

Having an AI strategy has become the norm. Knowing what it returns is still the exception. You probably carry that gap without seeing it. A budget committed, tools deployed, teams using them. Then someone asks for the return, and no one at the table can answer.

A figure that tells the truth about AI in business

The KPMG survey is clear. Across 2,110 leaders in 20 markets, in companies with at least 100 million dollars in revenue (three quarters of them above one billion dollars), 95 percent report having an AI strategy. It is now a given. No one is waiting anymore.

The second figure is more uncomfortable. Only 8 percent say they actually measure the return of that strategy. Not "hope for a return", not "are happy with it". Measure. With figures tied to a starting point.

Between the two sits a chasm. Almost every company has an AI intention. A tiny minority knows what it gets out of it. And this is not a question of technical maturity: same companies, same budgets, same tools. The difference lies in the discipline of measurement, not in the technology deployed.

The gap comes from framing, not from AI

The first reaction is to assume AI is not delivering. That is rarely the case. The problem almost always sits upstream.

An AI project starts on a broad intention: "get into AI", "not miss the shift", "automate". These wordings contain no indicator. They do not say which process, what gain is expected, or how it will be observed. Six months later the tool runs, teams use it, and finance asks for the return. Then, silence. Because there was no point of comparison defined at the start.

This is exactly what I detailed in the article on how to actually measure the ROI of an AI project: the return is not elusive, it is poorly measured because it is defined too late. The KPMG survey confirms that observation at the macro level. The gap from 95 to 8 is not a performance gap, it is a widespread framing gap.

Why 64 percent see benefits but only 8 percent prove them

The same report holds a detail that sums it all up. 64 percent of companies say they draw real benefits from AI. But only 8 percent measure them. How do you draw a benefit without being able to put a figure on it?

The answer is simple: most felt benefits are indirect. Teams move faster, decisions are better informed, the work is less tedious. All of that is true. But none of it enters a ROI sheet without prior measurement work.

The left column is felt. The right column is measured. Moving from one to the other does not depend on the AI tool, it depends on having set the indicator before starting.

Why 2026 no longer forgives the absence of measurement

For two years, the absence of measurement passed. The instruction was to move, to test, not to stand still while others jumped in. Budgets opened on the fear of missing out.

That period is ending. Finance teams now want tangible results. AI remains an investment priority, even in a downturn, according to the KPMG survey: budgets are not disappearing, they are being scrutinized. When a budget comes up for review and no one can show what it returns, it becomes an easy target for cuts.

In other words, not measuring is no longer just a lack of rigor. It has become a direct budget risk. The AI project you cannot defend with figures is the first one your leadership will question.

The real barriers to measurement

KPMG names the obstacles leaders themselves cite to explain why they do not measure. Three stand out.

• Lack of internal skills. Measuring AI ROI demands both technical and business skills, rarely found in one place.
• Difficulty quantifying indirect benefits. This is precisely the trap of the left column in the table above.
• Difficulty scaling. An isolated case that works only becomes measurable ROI once it deploys broadly, and many stay stuck at the test stage.

These three barriers share one thing: none is a technology problem. They are problems of method and framing. And that is good news, because method can be fixed.

How SolidScale closes this gap with the S3 method

The gap between strategy and measurement closes at the start, not at the end. In the Scan, Solve, Scale method, measurement is not a reporting step, it is the first question asked.

Scan: the free 30-minute audit does not only ask which process to automate. It asks which indicator you will judge the return on, and where the starting point sits today. A project that cannot be tied to any verifiable metric is a warning sign. Better to see it before committing a single euro.

Solve: metrics are set before the first line of code. The 4-to-8-week sprint delivers a usable tool. Each delivery is compared to the starting point measured during the Scan. You do not join the 64 percent who feel a benefit, you join the 8 percent who prove it.

Scale: monitoring includes a regular review of the indicators. Scope expands only when the figures justify it. You always keep a return you can defend in front of your finance team.

Key takeaways

The gap between 95 percent and 8 percent is not a technology problem. It is a method problem, and it is fixed at the start.

• 95 percent of companies have an AI strategy, 8 percent actually measure its return (KPMG, first quarter 2026).
• The gap comes from the fact that measurement is not set before deployment.
• 64 percent feel a benefit, but without an indicator defined at the outset, that benefit stays invisible in a ROI calculation.
• In 2026, not measuring has become a budget risk: finance teams now demand tangible results before renewing a budget.

The concrete action fits in one question: before continuing your AI strategy, define the indicator by which you will judge its return.

Frequently asked questions

What does "measuring AI ROI" really mean?

It means tying the return to a quantified indicator defined before starting, then comparing the current reading to that starting point. Feeling a benefit is not enough: until the gain is compared to an initial measure, it stays an impression, not a ROI.

Why do 95 percent have a strategy but only 8 percent measure the return?

Because strategy is decided in a meeting, but measurement is prepared upstream of the project. Most companies launch their AI initiatives without defining the return indicator at the start, which makes any later calculation impossible.

My company has an AI strategy, where do I start to measure?

With a single question before any new project: which figure will we judge the return on, and what is that figure today? Without an answer to that question, the project moves forward without a compass. That is exactly the starting point of a Scan audit.

Does measuring ROI slow down AI deployment?

No. Defining an indicator at the start takes a few hours of framing, not weeks. What really slows things down is deploying for months, then discovering you can prove nothing, and having to justify everything after the fact in front of a finance team.

Sources

• KPMG, Global AI Quarterly Pulse Survey: Q1 2026, kpmg.com/xx/en/our-insights/ai-and-technology/ai-pulse.html
• Siècle Digital, 95 percent of companies have an AI strategy, but only 8 percent truly know what it returns, 29 May 2026, siecledigital.fr
• Viuz, AI: 95 percent of companies have a strategy, but only 8 percent actually measure ROI (KPMG), viuz.com