CIOs Have Faced Gaps in AI Agent Oversight
A new survey found that most technology leaders lack the visibility needed to track AI agents across their organizations.
Updated on Sept. 24, 2026 in Artificial Intelligence
Live Poll
Do you trust that most companies are successfully managing their internal AI technology projects?
A Dataiku survey of 685 CIOs revealed significant failures in corporate AI governance, with 81% of leaders lacking visibility into agents built outside official channels. The findings highlight a growing divide between rapid employee-led adoption and formal IT oversight.
Why it matters
As 84% of CIOs report that employees create AI agents faster than they can govern them, the inability to manage these tools leaves companies vulnerable to performance failures and wasted investment. With 97% of leaders reporting board pressure to demonstrate returns, the lack of centralized control presents a major professional and operational risk.
Approximately 83% of surveyed organizations currently lack standardized lifecycle management for their AI agents, while 60% operate without a central governance layer. Additionally, 47% of CIOs reported decommissioning more than 20 individual agents in 2026.
The players
Dataiku
An enterprise AI platform provider headquartered in New York that offers software for data preparation, machine learning, and AI governance.
The Harris Poll
A global market research and analytics firm that conducts consumer and business surveys to provide insights on various industry trends.
Gartner
A prominent global research and advisory firm that provides insights and strategic guidance to leaders in IT, finance, and other business sectors.
The details
The report suggests that high-speed, decentralized agent creation is outpacing corporate oversight, leading to concerns about business outcomes and return on investment. To address these gaps, Dataiku has introduced new agent management software designed to centralize oversight and tracking.
Timeline
The Harris Poll surveyed CIOs between July 9 and July 29, 2026.
Dataiku released the survey findings on September 24, 2026.
New agent management software enters general sale in October 2026.
Companies face a deadline of the end of 2026 to hit AI performance targets.
Gartner expects 40% of agentic AI projects to be canceled by the end of 2027.
The Tech Race
This lack of governance marks a pivotal struggle between shadow IT and centralized infrastructure as organizations rush to scale autonomous AI. The industry currently faces a consolidation phase where companies must transition from experimental agents to strictly governed, high-output deployments.
Employees may face stricter oversight and centralized approvals for the AI tools they build or implement in their daily workflows. Consequently, users should expect a transition toward standardized company platforms as IT departments move to decommission unauthorized agents.
The takeaway
The rise of shadow AI agents creates a significant accountability gap that risks both budgets and professional reputations for technology leaders. Implementing centralized management is no longer optional for organizations aiming to avoid the mass decommissioning of projects expected by the end of 2027.
What happens next
General availability for the new Dataiku agent management software is scheduled for October 2026.
Further reading
For more information on the evolving standards for machine learning, visit our Artificial Intelligence section.
Live Poll
Do you trust that most companies are successfully managing their internal AI technology projects?
