56% of CEOs worldwide say they've seen no significant financial benefit from AI so far

PwC's 29th Global CEO Survey is precise about what "no significant financial benefit" means: neither higher revenue nor lower costs from AI over the past year (Source: PwC, 29th Global CEO Survey, 2026). Only 12% of the CEOs report both cost savings and revenue gains (PwC). Against that backdrop, CEO confidence in revenue growth has fallen to a five-year low, with just 30% strongly optimistic about the next 12 months, down from 38% last year (PwC).

The gap is between piloting and knowing#Link to this section

PwC found a growing divide between companies experimenting with AI and those deploying it at scale, and the survey ties returns directly to how deeply AI is embedded (PwC). CEOs reporting both cost and revenue gains are two to three times more likely to say they have embedded AI extensively across products, services, demand generation and strategic decision-making (PwC). Foundations matter just as much. Leaders whose organisations built strong AI foundations, such as responsible AI frameworks and technology that enables enterprise-wide integration, are three times more likely to report meaningful financial returns (PwC).

A separate MIT study reached a similar conclusion: 95% of the generative AI pilots it studied showed no measurable profit and loss impact (Source: MIT Project NANDA, The GenAI Divide: State of AI in Business 2025). It is worth reading that figure with care, since the study defined success narrowly, as measured profit and loss impact within about six months. But both studies point the same way: most organisations can't connect the tools they bought to a financial outcome, because they never set up the measurement to do it.

What good ROI measurement actually tracks#Link to this section

If you want a defensible answer on whether your AI costs are paying off, three layers need to be measured, not assumed.

Adoption at the tool level. Start with who actually uses each tool, how often, and for which tasks. This is where a lot of money leaks quietly. PwC found that even in the strongest use cases, only a minority deploy AI extensively: demand generation at 22%, support services at 20%, and product development at 19% (PwC). Good looks like a licence base where most seats are active weekly. Waste looks like paid seats nobody opens. You can't see the difference without measuring it per person, per tool.

Impact at the work level. Adoption alone is not value. Track time saved, output quality, and where the tool removed a bottleneck versus where it added review overhead. Ask the people doing the work, at short intervals, so the signal reflects reality rather than the launch-day demo. Integration overhead, rework and duplicated tools all show up when you measure impact properly, and stay invisible when you don't.

Financial outcome at the portfolio level. Roll adoption and impact into a simple view: cost per tool, estimated value returned, and net position. Read it as a range, not a single confident number, and refresh it as usage changes. Good looks like a small set of tools accounting for most of the value, and a clear list of the ones to cut. That matches the winners in the PwC data, who are two to three times more likely to have embedded AI extensively across their products, services, demand generation and strategic decisions (PwC).

Waste persists because most organisations measure none of these three, or measure the first and stop. Licences renew on autopilot. Pilots get praised in the boardroom and quietly stall in the field. And because the impact was never tracked, there is no defensible number to justify keeping the tool, or the confidence to drop it.

The CEOs reporting both cost and revenue gains had better foundations and deeper deployment, and both are states you can measure and move towards on purpose. That is the whole idea behind The GAiGE.