SAP survey finds AI ROI rising without clear cost savings

SAP said enterprise satisfaction with AI returns has reached 69%, based on a survey with Oxford Economics of 2,600 director-level and C-suite executives across midmarket and enterprise companies in 13 countries. The finding is useful for vendors selling AI into large accounts, but it also undercuts a common budget pitch: companies are reporting value, yet not primarily from lower costs or saved time.

The survey, released Wednesday, found that AI use has expanded since last year, while many organizations still have trouble deciding what counts as a return. Satisfaction with AI ROI rose by a few percentage points from the prior year, according to SAP, but nearly the same share of respondents said they were not persuaded that the technology they have deployed is producing its full value.

Sean Kask, SAP’s chief AI strategy officer, told CIO Dive that ROI is difficult for companies to measure and that organizations often weigh expected value against implementation effort. His framing points to a more prosaic phase for enterprise AI: less experimentation for its own sake, more pressure to choose use cases that can survive finance and governance review.

Value is showing up outside the cost line

SAP said respondents were more likely to see AI helping workers generate business insights, support decisions and interact with customers than to see it producing cost efficiency or productivity gains. Kask said cost savings remain part of the benefit case, but SAP’s survey did not find them to be the leading driver.

That distinction matters for enterprise software buyers and sellers. Many AI business cases have been built around headcount leverage, faster workflows or lower operating costs. The SAP data suggests that, at least for surveyed companies, the more visible returns are tied to decision support and customer-facing work, categories that can be harder to isolate in a clean ROI model.

Usage is still climbing. SAP and Oxford Economics found that companies now complete an average of 30% of tasks with AI assistance, up from one-quarter last year. Respondents expect that figure to reach 48% over the next two years, a projection that should be read as an expectation rather than a result.

Deployment maturity remains limited. Only 18% of companies reported using end-to-end, cross-functional AI deployments, compared with narrower single-task uses. That gap helps explain why ROI remains uneven: one-off tools can show local wins, while broader process change requires data access, controls and organizational redesign.

Spending is rising faster than proven returns

U.S. companies in the SAP survey said they spent an average of $37.2 million on AI this year and expect to raise that spending by 46% over the next two years. They reported $9.9 million in AI ROI this year and expect $26.5 million over the next two years.

Those figures show continued budget commitment, but they also leave open how companies define ROI and how much of the projected return depends on adoption that has not yet happened. SAP did not describe AI as a uniformly proven efficiency engine. Its findings instead show a market where usage and confidence are increasing before measurement practices have caught up.

A KPMG survey earlier this month found that leaders at more mature companies expect to focus more on scaling AI and determining its role in the business. Kask identified data, skills and governance as continuing obstacles to realizing returns.

Agentic AI adds another control problem. Kask told CIO Dive that companies rolling out agents are finding shadow agents and other unsanctioned tools with access to data or system actions that may not be auditable. For boards and operators, that turns AI ROI into a governance issue as much as a technology procurement question.

Kask said companies with board-level AI literacy and a willingness to change systems are better positioned to produce measurable outcomes. That is a higher bar than adding copilots across departments, and it is where the next phase of enterprise AI spending will be judged.

This story draws on original reporting from CIO Dive.