Enterprise AI ROI survey: most leaders see returns, few can measure them

Surveys on enterprise AI return tend to produce one headline. Battery Ventures' 2026 enterprise spend survey, published July 17, is more useful because it separates two questions: do companies see returns, and can they show them. The answer to the first is "some do." The answer to the second is "mostly not yet" ^[1].

Where the returns are

Returns are uneven across projects. In the survey, 31% of respondents said fewer than 25% of their AI projects have delivered measurable ROI, while 16% said more than half of their projects are delivering measurable returns. By function, 28% of CXOs named software development and engineering productivity as the area with the clearest return ^[1].

How companies measure it

Of those surveyed, 42% measure ROI in some areas but not consistently, and 43% are still defining how to measure it ^[1]. When they do measure, they lean on cost and time rather than revenue:

How to read it

Two cautions. First, the source page does not publish the sample size, respondent profile or fieldwork dates, so the percentages describe this survey's respondents, not enterprises in general ^[1]. Second, "clear ROI from some investments" is a low bar: it is compatible with most projects delivering nothing measurable, which is what the 31% figure above suggests. Surveys using different definitions of ROI will report very different numbers, so compare methodology before comparing percentages.

The figures above rest on a single survey, so treat them as one data point. A second 2026 survey asks a different question but points the same way on how ROI is judged. The Futurum Group's 1H 2026 Enterprise Software Decision Maker Survey, of 830 global IT decision-makers, found that hard financial measures (revenue growth and profitability combined) nearly doubled to 21.7% of primary ROI-metric responses, while productivity gains, still the leading metric, fell from 23.8% to 18.0% ^[2]. Futurum's release does not give fieldwork dates. Read together, Battery's respondents lean mostly on cost and time to measure ROI, while Futurum's respondents report a shift toward financial metrics as the primary measure; the two surveys use different questions and samples and are not directly comparable.

The consistent signal is the measurement gap. Cost savings and time saved are easier to count than revenue, and a company that only measures them will undercount some value and overcount others, such as hours "saved" that were never redeployed.

Practical implication

Before the next pilot starts, write down the one metric, its baseline and the review date. The survey's 6% with a consistent framework is where the advantage currently sits.