Dataiku report finds 83% of CIOs lack AI agent management rules
Survey of 685 CIOs in eight countries also finds 84% say IT cannot keep pace with staff building AI agents, according to Dataiku.
Dataiku published its "2026 Global AI Reality Report: CIO Edition" on October 18, a survey of 685 CIOs at large companies in eight countries, including South Korea, the US, UK and Japan, conducted with Harris Poll between July 9 and July 29.
83% of respondents said their company has no standardized framework for managing the lifecycle of AI agents, while 67% estimated that 51 or more AI agents are already running at their companies.
Most CIOs also reported difficulty measuring what those agents deliver: 72% overall, and 79% of South Korean respondents, said they cannot continuously measure AI agent operational performance against business outcomes. 84% overall, and 87% in South Korea, said IT departments cannot keep up with the pace of employees' AI agent development.
Board-level pressure is rising alongside. 97% said pressure to demonstrate measurable results from AI investment has increased, and 72% expect AI budgets to be cut or frozen if performance targets are not met by the end of this year.
Akamai flags data exposure in chatbot and agent use
Separately, Akamai's "Internet Security Status" report found sensitive information, including personal data, appeared in more than 6% of enterprise AI chatbot conversations. 47% of those conversations containing sensitive information took place through employees' personal accounts outside corporate monitoring, and more than 40% of enterprise users have installed AI browser extensions, a quarter of which changed their permission settings within 12 months of installation.
Akamai said the Model Context Protocol (MCP) that AI agents use to reach external systems is a blind spot in corporate security management, and recommended granting permissions by degree based on how well an agent's results can be verified and how recoverable failures are. Chief Security Officer Boaz Gelbord said three major threats are occurring at once: use of AI coding and productivity tools, direct AI integration into customer products and cloud workloads, and AI-based attacks targeting enterprises.
Quick answers
How many CIOs did Dataiku and Harris Poll survey?
685 CIOs at large companies in eight countries, including South Korea, the US, UK and Japan, between July 9 and July 29.
What share of companies lack an AI agent lifecycle management framework?
83% of respondents said their company has no standardized AI agent lifecycle management framework.
What did Akamai find about AI chatbot conversations?
Sensitive information, including personal data, appeared in more than 6% of enterprise AI chatbot conversations, and 47% of those took place through employees' personal accounts.