Data visibility remains a challenge

Autonomous AI systems in cybersecurity require reliable data and reference frameworks to enhance organizational resilience. Over the past decade, visibility has been a persistent challenge in enterprise IT, with teams struggling to secure, manage, and correct what they could not see.

Improved tools but slow action

This led to significant investment in monitoring and analysis tools, as well as dashboards, which have improved visibility. However, the main difficulty now is acting quickly based on the information available to reduce risks. This frustration is becoming more evident among companies, as the problem is no longer the lack of data but the inability to effectively exploit it when decisions need to be made quickly.

Structural issues in IT systems

Beyond data and visibility, the problem is primarily structural. Enterprise IT has developed around specialized tools, each performing its function correctly but none providing a single common and reliable reference framework. This results in scattered data, significantly slowing its exploitation.

Fragmented data slows responses

Questions that should be answered in a few seconds become a real treasure hunt between different systems, which sometimes describe the same machine in different ways. This fragmentation of IT and security data remains widespread in large organizations, turning quick corrections into long and complex interventions.

Patching becomes a strategic issue

The management of software patches (patching) offers one of the most telling examples. The number of vulnerabilities disclosed each month has now far exceeded what manual triage can reasonably absorb, while AI-assisted discovery only intensifies the pressure. What was once a routine maintenance operation now represents a significant risk for the company.

Strategic decisions in IT

Boards of directors and audit committees are directly interested in patch compliance, a topic that would have remained confined to IT teams only a few years ago. This is one of the new realities of enterprise IT: operational decisions are becoming strategic decisions. When the risk is that of a data breach, applying patches is no longer a simple IT task; it falls under enterprise-wide risk management.

Need for a reference framework

However, AI cannot autonomously determine which assets exist, to whom they belong, what has changed, or whether an action is compliant with the company's policies. These answers require the existence of a reliable reference framework. Facing fragmentation, the usual response is to add a new tool tasked with linking those already in place.

Integration as a principle

Too often, this approach adds a layer of complexity instead of removing one. On the contrary, the challenge is to make integration a starting principle. This implies consolidating the environment around a shared reference system: a single layer of trust capable of cataloging what exists in the environment, understanding how the various elements are connected, and knowing their current state.

Reliable data for AI success

This layer may be less visible than the latest AI feature, but it is precisely this layer that allows these new capabilities to be truly effective. This approach is part of a broader evolution in enterprise IT. Organizations must go beyond systems designed primarily to track tasks and processes to have reference frameworks capable of providing a reliable and authoritative context for decision-making.

AI depends on reliable context

It is precisely this context that modern AI systems depend on. It is also essential for IT and security teams that want to automate more of their operations. The reliability of an autonomous system directly depends on the reliability of the data on which it relies.

AI needs authoritative data

Many AI initiatives fail not because the model is insufficient, but because it works with incomplete, outdated, or contradictory information. AI can process information and formulate recommendations faster than any human team. However, it cannot know on its own which assets the organization possesses, understand the real state of its environment, or determine whether a situation complies with current requirements.

Governance is essential

Autonomy also requires governance. Any system capable of taking action must operate with clear safeguards: AI can handle routine tasks to which it is best suited, while humans retain oversight whenever judgment, responsibility, or risk assessment come into play.

Speed and control must balance

Speed matters, but risk exposure accelerates as well. However, speed without control only amplifies risks. Therefore, the future does not lie in autonomous systems operating without human intervention, but in governed autonomy, where AI acts within defined policies, based on reliable data, and within limits set by humans.

Converging pressures on IT

Three forces are converging today: the intensification of cyber threats, the pressure to demonstrate measurable returns on investments in AI, and the expectations of boards of directors, who demand more efficiency without additional complexity. Leaders can no longer consider security, operations, data, and AI as separate subjects.

Interconnected systems need a foundation

Everything is interconnected, and the foundation on which they rely must be as well. This does not mean eliminating existing tools or hoping for an overnight transformation. However, the direction to follow is clear, and the cost associated with maintaining fragmented environments is becoming increasingly difficult to justify.

Leaders must prioritize consolidation

For leaders, the starting point is therefore to give the impetus: consolidate the data, allow systems to safely take over routine tasks, and provide teams with the means to focus on problems that truly require human judgment.