Share with your CFO
SAP, surveying 2,600 director and C-suite executives across 13 countries alongside Oxford Economics, finds that AI satisfaction is climbing but the returns aren’t landing where the business case promised. The average U.S. company spent $37.2 million on AI this year and recovered $9.9 million in ROI, a ratio that’s hard to defend in a budget review. Cost efficiency and productivity, the two benefits most CFOs funded, ranked below insight generation and customer interaction as actual outcomes. Spending is projected to rise 46% over two years, with expected ROI jumping to $26.5 million.
What this means for your business
The math here is uncomfortable and worth sitting with. Companies are spending roughly four dollars to recover one, and the top benefits respondents are actually seeing, better decisions, richer customer interactions, don’t map cleanly to the efficiency narrative that got these budgets approved. If your organization’s AI business case was built on headcount reduction or productivity ratios, the gap between what was sold internally and what’s being delivered is already a governance problem, not just a measurement problem.
The SAP finding that only 18% of companies have deployed AI across end-to-end, cross-functional workflows matters here. Point solutions feel safe because they’re scoped and measurable, but they also cap the return. The ROI that compounds tends to live at process boundaries, where one AI-assisted step feeds another, and most organizations haven’t built the data infrastructure or governance frameworks to get there. SAP has an obvious interest in selling integrated platforms over point solutions, which puts a mild optimistic tilt on the cross-functional framing, but the underlying logic holds regardless of who benefits from it commercially.
The number to watch is the task-completion rate, currently 30% of all tasks completed with AI assistance, expected to reach 48% in two years. If that projection lands anywhere close, the ROI denominator grows faster than most current measurement frameworks can track. The CFOs who will defend their AI budgets in the next planning cycle are the ones building the measurement architecture now, not the ones waiting for the platform to report outcomes automatically. Shadow agents, ungoverned workflows, and unauditable system actions are already surfacing as deployment scales, and those are liability items before they’re ROI items.
Based on reporting from AI ROI is rising, but not where companies expected, originally published 2026-07-15 03:00:00.
