On June 8, 2026, the IBM Institute for Business Value released a tech leader study, conducted with Oxford Economics from January to April 2026, surveying 2,000 technology executives across 33 geographies and 19 industries. An earlier draft of this article misstated the "two-thirds" figure as AI pilots failing to reach production; the first-party source actually says about two-thirds of surveyed CIOs/CTOs are accountable for AI systems they cannot fully control. It must not be used as a project failure rate.
What the study shows is that control and governance capabilities are lagging behind deployment speed: 70% of respondents say business teams deploy technology faster than IT can track it, only 11% consider themselves fully prepared for the scale of AI agent deployment over the next year, and 77% say AI adoption has outpaced their existing governance capabilities. These figures reflect respondents' views and do not mean every company will face problems in the same proportions.
Scaling risk does not come only from model performance. 59% of surveyed technology executives cite security and compliance as the main barriers to scaling AI agents; 84% have not yet fully integrated AI financial management into operations, and 85% still lack full visibility into real-time AI spending. If you cannot answer at the same time who owns a system, what it can access, how to stop it when an incident occurs, and whether costs still align with business objectives, then a pilot passing technical testing does not mean it is ready for production operations.
From the perspective of "The CIO for Your Business Operations," Unima recommends putting four controls in place before scaling: 1. build an inventory of AI systems and agents, recording business owners, data sources, and vendors; 2. restrict read/write permissions based on risk and retain human approval for high-impact actions; 3. log exceptions, security incidents, response times, and stop/rollback procedures; 4. track usage, cost per task, human rework, and business outcomes together. If any item lacks an owner or a verifiable threshold, narrow the scope first.
This checklist is an implementation recommendation based on IBM's first-party research. It does not represent specific client results, nor does it guarantee that every AI project will scale successfully. Companies must still set controls according to data sensitivity, use cases, jurisdiction, vendor contracts, and internal risk tolerance.
Official sources: IBM Newsroom, "New IBM Study Finds CIOs and CTOs Face Growing AI Control Gap as Enterprise Deployment Scales", 2026-06-08, https://newsroom.ibm.com/2026-06-08-new-ibm-study-finds-cios-and-ctos-face-growing-ai-control-gap-as-enterprise-deployment-scales; IBM IBV, "2026 Tech Leader Study: Building the IT foundation for agentic AI at scale", https://www.ibm.com/thought-leadership/institute-business-value/en-us/c-suite-study/cxo. This article was fact-checked on 2026-08-24.