Rising AI Spend, Elusive Returns: What KPMG's Data Says About Measurement

KPMG’s latest Global AI Pulse lands on the exact problem we built The GAiGE to solve. Spending is holding, usage is climbing, and yet almost nobody can prove the return. The interesting part is what separates the few who can.

The headline numbers set the scene. AI remains a top investment priority for 79 percent of leaders, up from 74 percent the previous quarter, with average AI spending holding steady at 188 million dollars (Source: KPMG Global AI Pulse press release, June 2026). Everyday use is rising fast too, with 22 percent of organisations now describing AI as part of everyday work, up from 13 percent, the largest single-quarter shift on KPMG’s maturity curve (Source: UC Today, June 2026). But momentum is not the same as return. Only 7 percent of leaders report establishing ROI, even as nearly one in four face pressure to prove value to investors (Source: KPMG Global AI Pulse press release, June 2026).

So spend is up, usage is up, and proof is scarce. The gap between activity and value is the whole story.

The real predictor is cost visibility, not tool choice

Here is the finding that matters most for anyone signing off budget. Leaders with strong cost visibility are five times more likely to report established ROI, 15 percent against 3 percent (Source: KPMG press release, June 2026). Yet only about a third of respondents report full visibility into AI operating expenses, and 42 percent have only partial visibility into AI spending (Source: Techstrong.ai, July 2026; consulting.us, June 2026).

The consequences are already showing up in cancelled work. Around 33 percent cite limited understanding of AI cost structures, including how token-based pricing works, as a major challenge, and nearly half have already scaled back AI agent deployments because costs outweighed the benefits (Source: UC Today, June 2026).

If you want to move from the 3 percent group to the 15 percent group, this is what to track. Start with cost per tool per team, not a single line item for AI. Track spend against actual usage, so a licence nobody opens is visible as waste rather than hidden in a subscription total. Watch the trend on usage-based charges, since token consumption moves with behaviour and can drift well past forecast. What good looks like is simple to state: you can name, for any AI tool, what it costs, who uses it, and what changed in their work because of it. Most organisations cannot do that yet, and the data shows it.

Executive ownership is the other half

Cost visibility does not appear by accident. It follows from someone owning the outcome. KPMG found that only 24 percent of leaders say the CEO is ultimately accountable for AI-driven business outcomes, while 29 percent point to the broader C-suite, which in practice can mean accountability belongs to everyone and no one (Source: UC Today, June 2026).

Where ownership is clear, results follow. Organisations where the CEO is explicitly accountable report higher confidence in their AI strategy, 60 percent against 22 percent, are more likely to realise meaningful business value, 57 percent against 21 percent, and are nearly four times more likely to report established ROI, 14 percent against 4 percent (Source: UC Today, June 2026).

The measurement lesson is that ownership needs a number attached to it. An accountable executive without a live view of adoption and cost is accountable for a guess. So pair the name with a cadence. Decide who reviews AI cost and impact, how often, and against what threshold. More than half of leaders now report having AI cost monitoring dashboards in place, and a similar share have embedded cost reviews into their AI approval processes (Source: UC Today, June 2026). That is the discipline the leaders are using, and it is repeatable.

The missing layer between activity and value

What KPMG describes, and what we see, is a gap that spreadsheets alone do not close. Finance can see the invoice. IT can see the licence count. Neither of those tells you whether a tool actually changed how work gets done, or whether people quietly went back to the old way.

That is the measurement layer we focus on. Short anonymous micro-surveys ask the people using the tools whether they save time, where the tool helps, and where it does not. Read alongside spend and adoption, that turns AI from a line of cost into a line you can defend. When someone asks whether the AI budget is working, the answer stops being a story and becomes a number with evidence behind it.

The organisations pulling ahead are not the ones spending the most or using AI the most. They are the ones who can see clearly and who put a name on the outcome.

If that is the number you are missing, start a GAiGE trial and see it for your own teams within a week. We would rather show you than tell you.