IBM’s survey depicts a divide inside companies using AI: agents are acting across workflows, while few surveyed leaders control the investment decisions that shape where those systems are deployed.
IBM and Oxford Economics gathered responses from 1,667 senior executives across 33 geographies and 20 industries from April to June 2026. Respondents were responsible for AI strategy, governance, implementation or value realization. They included chief AI officers, chief data and AI officers, heads of AI and equivalent leaders, which IBM groups under the label “CAIOs.”
Nearly half of respondents, 48%, said AI agents can start and complete tasks spanning workflows. At the same time, 75% said governance, risk or compliance delayed AI-driven action. While 73% said their organizations had identified decisions requiring human oversight, 61% said AI decision rights were clearly set out and applied consistently.
Only 18% reported full control over AI investment decisions, and 15% said they had full authority over AI workforce strategy. When performance measures point to a chance to improve results, 29% said their organizations quickly adjust decisions and resources. Another 40% said decisions are discussed but action is delayed.
IBM said weak visibility into AI costs and difficulty connecting specific investments to business results can leave funding and talent with initiatives that no longer warrant investment, while more promising work struggles to expand.
In the group IBM classified as high-impact, respondents put the share of AI projects producing measurable business results at 63%, compared with 43% among other respondents. In that group, 94% said business outcomes inform decisions about additional funding and talent. Another 96% said they stop or reduce investment in initiatives that fail to deliver, compared with 46% of other respondents.
IBM’s analysis found associations between management practices and reported results, but did not establish that those practices caused the differences.



