Argility and Smollan expose the AI confidence divide in South Africa

The Growing AI Confidence Gap: Why IT Leaders Are Losing Faith

In recent months, the landscape of artificial intelligence (AI) in enterprise environments has shifted dramatically. What was once a wave of optimism and rapid adoption is now giving way to a growing sense of uncertainty among IT leaders. According to the Argility Technology Group (ATG), a part of the global entity Smollan, there is a clear and widening gap between the capabilities of AI systems and the confidence that enterprises have in managing them effectively.

CJ Oosthuizen, a Google Cloud and Workspace specialist at ATG, explains that just six months ago, many IT leaders were confident in their AI deployments, with 40% declaring their efforts as “mature.” However, this confidence has since collapsed, with only 23% of IT leaders now considering their AI implementations mature.

The Paradox of More AI, Less Confidence

Oosthuizen points to a report from JumpCloud for Q3, which highlights the current state of AI adoption in enterprises. While the number of AI deployments has increased, the confidence in how these systems are being managed has dropped significantly. This paradox reflects a critical issue: even though more AI is in use than ever before, there is a lack of visibility and control over how it operates.

The report suggests that the problem lies not in the speed of AI adoption, but in the infrastructure and governance that support it. As AI agents gain deeper access to enterprise data and execute tasks autonomously, many organizations are realizing that their existing systems are not equipped to handle this new level of complexity.

Agents in Production: A New Reality

One of the most striking findings from the report is the shift from passive chatbots to active AI agents. More than 60% of organizations are running AI agents in production workflows, and the trend is accelerating. These agents are not just generating text—they are executing tasks, making decisions, and manipulating data without human intervention.

Worse still, full AI autonomy without human review has more than doubled in a short period, jumping from 11% to 26%. This means that a quarter of enterprise AI implementations are now operating entirely on their own, raising serious concerns about accountability and oversight.

The Rise of Non-Human Identities

The proliferation of AI agents has also led to a significant change in enterprise network perimeters. In 83% of organizations, non-human identities (NHIs)—such as API keys, service accounts, bots, and AI agents—now outnumber human users. Despite this, only 21% of organizations have established governance controls for these identities.

This lack of oversight means that the majority of machine identities operating within enterprise systems are unmonitored, unmanaged, and unaccountable. Oosthuizen emphasizes that this is a major risk, as these identities can potentially cause harm if left unchecked.

The Risk Velocity Problem

When autonomous execution is combined with a total lack of identity governance, the risk of security breaches and operational failures escalates exponentially. The report reveals that three in four IT leaders (75%) admit that AI is advancing faster than their ability to manage the associated risks.

This situation is particularly concerning because traditional IT tools and processes are not designed to handle the complexities introduced by AI. For example, when a human employee leaves an organization, deprovisioning their access is a standard workflow. But when an AI agent is hardcoded into multiple systems via a legacy service account, revoking its permissions without disrupting business operations becomes a challenge.

Reclaiming Control: The Path Forward

To address the AI confidence gap, the report suggests that IT leaders must shift their focus from the AI applications themselves to the identity and infrastructure layers that support them. Rather than stopping innovation or banning AI tools, the solution lies in building robust governance frameworks that can keep pace with the evolving AI landscape.

Key recommendations include:

• Unify human and machine identity: Machine identities should no longer be treated as secondary configuration details. They must be brought into the central directory, subjected to the same strict zero trust access controls, and continuously monitored.
• Implement just-in-time (JIT) privileges for AI: Autonomous agents should not have permanent administrative access. Instead, temporary, tightly scoped permissions should be granted and expire immediately after a specific workflow is executed.
• Bridge the visibility gap: IT teams need a single control plane that can identify every AI agent, understand what data it has access to, and track its actions in real-time.

Conclusion

As the initial rush of the AI boom gives way to a more realistic understanding of its implications, IT leaders must adapt their strategies to ensure that AI is both powerful and secure. By investing in unified, identity-centric infrastructure, organizations can close the risk gap and turn a volatile environment into a competitive advantage.

Share :