Drop in AI Confidence: A Sign of Business Maturity

Confidence in artificial intelligence is falling, and that's excellent news. According to JumpCloud's Q3 2026 IT trends report, the percentage of IT leaders who consider their organizations mature in AI deployment has dropped from 40% to 23% in just six months. At first glance, it might be interpreted as a setback, but the data tells a more nuanced story: organizations losing confidence are precisely those that have moved their AI agents from pilots to production.

From Euphoria to Realism

The study, which surveyed 800 IT leaders in the US and UK, shows that 84% of organizations plan to expand AI use in IT operations over the next 6 to 24 months. The drop in confidence is not a retreat but an adjustment to the reality of what production entails. In a pilot, an AI agent does one thing in a controlled environment. In production, it accesses real systems, makes decisions that affect real workflows, and operates continuously, often without human oversight. This phenomenon echoes the 'trough of disillusionment' in Gartner's hype cycle, where after the peak of inflated expectations, technology faces a phase of realistic adjustment. In 2024, a similar McKinsey study showed that only 8% of companies had scaled AI to production level, while in 2026 that number may be higher, but with harsher lessons.

“Organizations losing confidence in AI are the ones most likely to get it right,” the report notes.

The JumpCloud survey reveals that companies reporting a drop in confidence are those that have deployed agents in production and are facing real governance, security, and scalability issues. This is not an abandonment of AI but a learning process. For example, in the financial sector, several banks that launched chatbots to the public in 2023 had to withdraw them due to costly errors, but later relaunched them with stricter controls. This cycle of 'deploy, problem, adjust' is now the norm.

The Governance Gap

The hardest problem in enterprise AI today is not capability but accountability. The key data point: non-human identity governance is the least adopted AI security practice, present in only 21% of organizations. Non-human identities (AI agents, service accounts, etc.) already outnumber human users in 83% of organizations, and that population is growing fast. Yet most of these identities lack the governance structures every human employee has: no formal registration, no designated owner, no defined access scope, no deactivation process. These zombie agents keep running, accumulating permissions, and accessing systems without control. The gap between deployment and control is where risk accumulates.

To put it in perspective, in 2023, an error in an AI agent at a logistics company caused a 12-hour supply chain disruption because the agent had excessive permissions to modify databases. Such incidents multiply as more agents enter production. The JumpCloud report highlights that only 21% of organizations have governance practices for non-human identities, meaning the remaining 79% operate with significant risk. Compared to human identity governance, which has over 90% adoption in mid-sized companies, this gap is alarming.

What Are the Consequences?

Organizations that have closed this gap share characteristics: consolidated IT environments, AI agents treated as governed identities, and measurement of real outcomes. Those at the top maturity level are five times more likely to report no barriers to expanding their AI agents. They are not more cautious about AI but more confident, because they built the foundation that makes trust earned, not assumed. For the reader, this means the drop in confidence is not a sign of failure but of learning. Companies that are honest about production challenges will be better positioned to scale AI safely and effectively.

In the market, this trend is driving demand for non-human identity management solutions. Startups like Oort and ConductorOne have seen a 300% increase in inquiries since 2025. Additionally, cloud providers like AWS and Azure are integrating AI governance tools into their platforms. For users, this translates to greater transparency and control over agents interacting with their data. For example, an IT admin can now see which agent accessed which file and when—unthinkable two years ago.

However, the path is not uniform. SMEs face greater challenges because they lack resources to implement these practices. According to the report, companies with over 500 employees are twice as likely to have non-human identity governance as smaller ones. This could create a digital divide in safe AI adoption, where large companies scale while small ones stagnate or take excessive risks.

What Readers Should Know

• AI confidence is dropping because companies are moving from pilots to production and encountering real governance issues.
• Non-human identity governance is the weakest point, with only 21% adoption.
• Organizations that consolidate their environment and treat AI agents as governed identities achieve better results.
• This drop is a sign of maturity, not regression.
• The AI security market is growing rapidly, with startups and major providers competing to offer solutions.
• SMEs are at a disadvantage, which could widen the safe innovation gap.