Only 13 per cent of IT application leaders say they have appropriate governance structures to manage AI agents, according to a Gartner survey that highlights significant readiness gaps as organisations rush to deploy artificial intelligence technologies.

The research found just 15% of IT application leaders are considering, piloting or deploying fully autonomous AI agents - systems that operate without human oversight. This contrasts sharply with the 75% that are piloting or deploying some form of AI agents, suggesting organisations favour supervised implementations over autonomous systems.

In May and June 2025, Gartner conducted an industry-wide survey of 360 IT application leaders from organizations with at least 250 full-time employees in North America, Europe and Asia/Pacific, with the aim of understanding the impact of generative AI (GenAI) and agentic AI across enterprise applications.

“The hype around agentic AI continues to grow, with vendors positioning AI agents as the next phase of AI evolution that will address the shortfalls of more traditional GenAI assistants,” said Max Goss, Senior Director Analyst at Gartner.

Security concerns present a substantial obstacle, with 74 per cent of respondents viewing AI agents as a new attack vector. Only 19 per cent expressed high or complete trust in vendors' abilities to provide adequate hallucination protection - a critical issue where AI systems generate false or misleading information.

The governance and security challenges are compounded by a lack of organisational readiness and trust in vendor security capabilities, according to the survey of 360 IT application leaders from organisations with at least 250 employees across North America, Europe and Asia-Pacific.

Productivity expectations remain modest

While 26 per cent of respondents anticipate transformative productivity impacts from AI agents, most (53 per cent) expect significant but not transformative results. Twenty per cent predict only marginal gains.

Notably, organisations with strong alignment between IT, business users and leadership on AI problem-solving are 1.6 times more likely to expect transformative outcomes. However, only 14 per cent of respondents reported having this alignment—a gap that may explain divergent expectations about AI's value.

The survey suggests many organisations are targeting AI agents at suboptimal use cases. Analytics and business intelligence topped the list of high-impact domains (64 per cent), followed by customer service (55 per cent) and office productivity (39 per cent).

However, organisations lacking clear understanding of AI's business applications were nearly twice as likely to focus on office productivity. Goss noted office productivity "are not necessarily the areas that will provide organisations with the most value" for those without a strong AI strategy.

Despite widespread discussion about AI displacing workers and applications, survey respondents showed scepticism about near-term replacement. Only 12 per cent strongly agreed AI agents would replace applications within two to four years, while just seven per cent strongly agreed agents would replace workers in that timeframe.

However, when including those who somewhat agreed, 34 per cent believe agents will replace applications and 29 per cent expect worker replacement within two to four years - figures Goss described as "significant for technology that has only been generally available for the last 12 months."

Gartner recommends organisations develop platform-agnostic AI agent governance frameworks to reduce sprawl risks and provide clear guidelines and controls. The firm also advises targeting agents at high-impact domains rather than default office productivity deployments, and adopting multivendor strategies rather than relying on single vendors.