Confidence in enterprise AI deployments has dropped sharply in six months, showing a widening gap between adoption speed and risk management capabilities.
From optimism to uncertainty
In early 2024, 40% of IT leaders considered their AI implementations mature. That share has since fallen to 23%, even as more tools enter use. The change reflects growing awareness that supporting infrastructure hasn’t matched the pace or independence of these systems.
CJ Oosthuizen, a Google Cloud and Workspace specialist at Argility Technology Group, describes the situation as contradictory. “More AI is in active use, yet confidence in its implementation has declined,” he said. “Reality has caught up with the hype.”
A recent IT trends report from JumpCloud shows the disconnect. While organizations deploy AI agents quickly, many lack governance to track or control them. The result is a workforce of non-human identities—API keys, service accounts, bots—operating without oversight.
Autonomy without oversight
Over 60% of companies now run AI agents in production workflows, moving from simple chatbots to systems that act independently. Full autonomy, where AI operates without human review, has risen from 11% to 26% in recent months.
The increase has outpaced traditional IT controls. In 83% of organizations, non-human identities now exceed human users. Yet only 21% have established governance for these machine identities. Oosthuizen noted the risks: “Most identities inside enterprise systems remain unmonitored. When you combine autonomous execution with no identity governance, problems multiply quickly.”
Related: Rimini Street Posts Quarterly Financial Results
Three-quarters of IT leaders admit AI is advancing faster than their risk management abilities. The issue isn’t just technical but structural. Removing access for a human employee is straightforward, but revoking permissions for an AI agent embedded in multiple systems can break workflows. Few companies have a clear process.
The decline in perceived AI maturity isn’t a failure. The report explains it as a necessary adjustment. Six months ago, “maturity” meant deploying a chatbot or API. Now, leaders see it requires lifecycle management, data lineage, and identity security—areas where existing tools often fall short.
The shift may feel like a step back, but it could lead to better practices. The first wave of AI adoption focused on capabilities. The next phase will define boundaries and accountability. Without addressing infrastructure gaps now, risks will grow as AI becomes more integrated into daily operations.
Closing the gap
The report argues against slowing AI adoption. Demand will always find ways around restrictions. Instead, IT leaders must strengthen the identity and infrastructure layers that support these tools.
Oosthuizen recommends three steps: unifying human and machine identities under zero-trust controls, using just-in-time privileges for AI agents, and creating a single control plane to monitor autonomous actions. “By shifting focus from new features to infrastructure, IT leaders can turn instability into an advantage,” he said.
The challenge extends beyond technology. Companies that treated AI as a simple solution now face the reality that these systems need the same rigor as other critical infrastructure. The transformation is already happening—the question is whether businesses are prepared for it.
