A recent global survey of 685 chief information officers reveals that most lack comprehensive oversight of AI agents constructed outside approved IT systems, underscoring a widespread challenge in managing AI at scale.
- 81% of CIOs cannot fully oversee AI agents built outside formal IT systems.
- 83% lack standardized lifecycle management for AI agents across their organizations.
- Most CIOs face intensified scrutiny from boards and risk significant budget cuts if AI projects underperform.
What happened
Dataiku released the Global AI Confessions Report: CIO Edition, based on a survey of 685 CIOs across the US, UK, France, Germany, UAE, Japan, South Korea, and Singapore. The survey found that 81% of CIOs do not have complete oversight of AI agents created outside official IT channels. While 90% report full tracking of agents in their environment, 72% cannot confirm consistent delivery of intended business outcomes by those agents.
Further findings reveal that 67% of CIOs estimate more than 50 AI agents in production, but 83% lack standardized lifecycle management processes. Nearly half of the respondents have already decommissioned over 20 agents this year. Additionally, only 21% have near-real-time visibility into AI expenditure by team or use case, reflecting significant gaps in monitoring and governance.
Why it matters
The rapid creation of AI agents by employees outside the IT department outpaces CIOs’ ability to govern them, with 84% of CIOs agreeing this is a major challenge. This raises significant risks including uncontrolled costs, unclear accountability, and potential security or compliance issues. The survey revealed a fragmented view on responsibility when agents fail, distributing ownership among IT, data/AI teams, and security functions.
These governance gaps are critical as AI initiatives face mounting pressure from corporate leadership. Nearly all CIOs report increased board expectations to demonstrate clear returns on AI investments. Moreover, 87% say top executives have linked their job security directly to successful AI outcomes, with 76% fearing their role could be jeopardized if AI-driven gains do not materialize by the end of 2027.
What to watch next
Dataiku has introduced a new Agent Management solution that aggregates AI agents from multiple platforms to provide centralized visibility and control. The tool integrates with major cloud and AI service providers and will be available for general sale in October, signaling a growing market for AI governance software to help CIOs regain oversight.
Looking ahead, it will be important to monitor how organizations implement lifecycle management and governance frameworks for AI agents. Gartner forecasted in 2025 that over 40% of agentic AI projects will be cancelled by 2027 due to costs, questionable business value, and lack of risk controls. CIOs’ approaches to managing AI investments, including diversifying models and exploring open-source options, may shape their ability to sustain AI initiatives under evolving board scrutiny.