A new KPMG report finds that just 7% of business leaders can demonstrate a return on their artificial intelligence investments. The rest are pouring money into AI without being able to show it actually pays off.
The gap between belief and proof
The report paints a stark picture: most executives are convinced AI is creating value, but they can't back that conviction with numbers. That disconnect is more than an accounting problem. It's a strategic one. If a company can't measure what an AI system returns, it can't decide where to double down, where to cut, or even whether the whole effort is worth it.
KPMG's finding suggests the perceived value of AI is running far ahead of the hard evidence. Executives may be hearing success stories from peers, seeing flashy demos, and feeling pressure to keep up. But when asked to show the actual return on a specific AI project, most come up empty.
Why cost tracking falls short
Part of the issue is that AI costs are often scattered. There's the obvious spend on software and cloud compute, but also the hidden outlays: data cleaning, integration work, retraining staff, and the time engineers spend tweaking models. Many finance teams don't have a clear line on those expenses, so they can't match them against the benefits AI is supposed to generate.
Without that tracking, accountability gets fuzzy. If an AI system doesn't deliver, who's responsible? The team that built it? The vendor that sold it? The executives who approved it? The report implies that the lack of rigorous cost tracking makes it nearly impossible to assign responsibility or learn from failures.
What the report signals
The KPMG report doesn't offer a magic formula, but it does point to a clear need: better measurement and more disciplined oversight of AI spending. That means building cost-tracking systems that can follow a dollar from the initial investment to the eventual outcome, and then asking hard questions about whether the outcome was worth it.
It also suggests that companies should treat AI like any other capital investment. You wouldn't buy a factory without knowing its expected output. AI projects deserve the same scrutiny. The 7% figure is a warning that most organizations are flying blind on some of their most expensive bets.
The question now is whether finance chiefs will take that warning seriously. The report makes one thing plain: the gap between perceived value and proven return won't close on its own. It will take deliberate work on cost tracking and accountability.
