Just 18% of surveyed AI leaders have full investment control, IBM finds
The survey of 1,667 executives links stronger reported outcomes with performance-based funding decisions—including cuts to projects backed by senior leaders.
Loading page…
The survey of 1,667 executives links stronger reported outcomes with performance-based funding decisions—including cuts to projects backed by senior leaders.
Listen to this story
A survey by IBM and Oxford Economics of 1,667 senior executives across 33 geographies finds a control problem beyond deployment: 48% say agents already initiate and execute workflow actions, while 75% report governance, risk or compliance delays. IBM’s high-impact cohort reports more measurable project outcomes and faster enterprise scaling than other respondents. Those gaps point to investment discipline and the ability to intervene as important operating differences, though the survey is self-reported and does not establish that these practices cause stronger results.
Forty percent say performance insights prompt discussion but action is delayed, suggesting that having data does not ensure resources or decisions change.
Although 73% specify where human oversight is mandatory, only 61% say AI decision rights are clear and consistently applied.
High-impact leaders report measurable outcomes in 63% of AI projects, versus 43% for others; their projects reach enterprise scale in 8 months rather than 13.5.
Only 18% of enterprise AI leaders surveyed by IBM say they fully control AI investment decisions. Just 29% say their organizations quickly shift decisions and resources when performance data points to better results. The new study puts a concrete limit on AI leadership: responsibility for results does not necessarily come with authority to act.
Published November 16, 2026, A template for high-impact AI comes from the IBM Institute for Business Value, working with Oxford Economics. The researchers surveyed 1,667 senior executives between April and June 2026, across 33 geographies and 20 industries. Their questions covered AI priorities, investment management and putting AI into operation.
IBM calls the respondents chief AI officers, or CAIOs, but the sample extends beyond that title. It includes chief data and AI officers, heads of AI and equivalent executives responsible for strategy, governance, implementation and delivering value.
The investment findings distinguish recognizing an opportunity from acting on it. Forty percent of respondents say performance insights prompt discussion, but action is delayed. That is a different problem from having no performance information: the organization has identified a possible improvement without promptly changing its decisions or resources.
In the study’s foreword, IBM transformation and operations chief Joanne Wright describes AI leaders as having a broad view of where the technology creates value. Yet their results often depend on other people’s decisions. Her prescription goes beyond technical oversight: gather evidence, influence decisions and measure success.
The governance findings show another tension. Forty-eight percent say AI agents initiate and carry out actions across workflows. At the same time, 75% say governance, risk or compliance delays AI-driven action. The survey records both expanding operational autonomy and friction around its use—not simply a lack of AI deployment.
Seventy-three percent say their organizations have specified where human oversight is mandatory. A smaller share, 61%, say AI decision rights—the rules about who can decide what—are clear and consistently applied. Setting a boundary for human involvement and assigning decision-making authority are distinct measures of control.
IBM’s identified group of high-impact leaders reports more autonomy alongside stronger intervention capabilities. Some 96% say AI outputs can trigger actions autonomously, versus 40% of other respondents. They are also nearly four times as likely to say they can quickly pause, adjust or redirect AI systems without disrupting operations.
IBM associates stronger results with differences in investment discipline, workflow design and control over AI. These are executives’ reported outcomes and organizational practices, not proof that any single practice caused the performance gap. Within that comparison, high-impact leaders describe several concrete ways of moving money and people:
Across all respondents, only 31% say clear thresholds for stopping AI initiatives are critical to success. The contrast concerns more than choosing promising projects: the high-impact group reports a willingness to withdraw support after an initiative has already attracted resources and executive backing.
The ambition for broader change remains uneven. Thirty-two percent say AI is reshaping their business model, including what they sell, how they price or how they capture value. Another 35% say it primarily strengthens the existing model; 26% expect a reshaping but have not made the necessary shifts.
Asked where change should begin, AI leaders rank workflows and business processes first. Their stated priority is changing how work runs—not only selecting the next AI investment.
Loading discussion...
Join the conversation
Explain who should have the final say when results disappoint.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.