At ApiDays, one thing became clear across keynotes, talks, and use cases:

Artificial Intelligence is no longer a possibility to evaluate, but a reality that is already entering enterprise systems. The point, however, is not AI itself. The point is how AI can act in the real world. And this is where a key message emerged, one that ApiShare brought to the event stage: AI is no longer an option. APIs are the foundation.

ApiDays as an observatory: two talks, one shared message

ApiShare’s presence at ApiDays revolved around two different contributions, deeply aligned in their message.

On one side, a product-focused talk centered on a very specific question: what happens when AI agents become operational actors and start discovering, combining, and executing services through APIs?

On the other, the Fastweb + Vodafone talk offered a concrete enterprise perspective, showing how in complex ecosystems API governance is already a structural necessity, not an optional choice.

This perspective emerged particularly clearly in the talk by Tullio Nardi (Fastweb + Vodafone) at ApiDays, where API management was described not as a theoretical exercise, but as an essential lever to maintain control, consistency, and security in large-scale architectures (full talk available here:

Two different viewpoints, one conclusion: without governed APIs, AI does not scale.

AI agents: from assistants to execution systems

The real paradigm shift is not simply the spread of Artificial Intelligence, but the role AI agents are taking within digital systems.

For years, we have talked about AI as support: tools capable of suggesting, analyzing, and optimizing decisions made elsewhere. Today, however, the scenario has changed. AI agents are becoming operational actors, able to interact directly with the application ecosystem.

They no longer just provide recommendations. They act.

They query services, orchestrate complex processes, make autonomous decisions, and execute actions on real systems, often in real time. This shift from assistance to execution marks a profound discontinuity: AI is no longer just intelligence “in support,” but intelligence embedded within operational flows.

To function in this way, however, AI agents need a solid and reliable foundation. And this is where APIs come into play.

Agents learn through APIs, by observing and understanding the capabilities exposed by systems. And it is through APIs that they execute, invoking services, combining functionalities, and triggering processes.

In this sense, APIs stop being a simple technical integration layer and become the pillars of the new digital ecosystem: the point of contact between artificial intelligence, automation, and operational reality.

This is exactly the perspective shift at the core of the talk by Giorgio Sismanidis and Andrea Cinquepalmi at ApiDays: when AI agents start discovering, understanding, and using APIs autonomously, governance stops being an infrastructural concern and becomes an enabling component of the system’s intelligence itself (full talk available here:

The real risk: AI grows faster than governance

This is where one of the strongest tensions many organizations are experiencing today emerges, often without being explicitly named.

On one side, innovation accelerates. APIs multiply, teams become increasingly autonomous, architectures distribute and fragment. On the other, the first AI agents begin to appear, interacting with increasingly heterogeneous technological environments, combining services, data, and processes.

The problem is not this evolution in itself.

The problem is when it happens without a governance model.

In the absence of clear rules, shared visibility, and defined responsibilities, the risk is not only technical. It is systemic. Knowledge of what exists and how it is used is lost, control over flows weakens, inconsistencies multiply, and operational and security vulnerabilities become inevitable.

In this scenario, AI does nothing but amplify the problem. Because intelligent agents scale quickly, replicate behaviors, and interact with systems at a speed that makes any manual or informal approach simply unsustainable.

The Fastweb + Vodafone talk at ApiDays made this point very clear: when technological complexity grows faster than the ability to govern it, innovation stops being an advantage and becomes fragile.

And it is precisely in this space that governance stops being an organizational choice and becomes a strategic necessity.

Governing today, not tomorrow

One of the most common mistakes when talking about Artificial Intelligence is thinking that governance can wait. That it is something to introduce later, when AI is more mature, when agents are standardized, or when the technological ecosystem has stabilized.

At ApiDays, a very different and much clearer message emerged: governance must precede scale, not chase it.

The reason is simple. AI agents do not just use what they find: they amplify everything. They amplify inconsistencies, ambiguities, and design shortcuts. An undocumented API, an unmanaged version, or an untracked access do not remain isolated issues: they become risk multipliers when embedded into automated and autonomous flows.

In this context, postponing governance does not mean gaining speed. It means accumulating structural debt that AI will make visible and costly, very quickly.

Governing today means making a strategic choice: building the digital fabric on which Artificial Intelligence can operate in a reliable, secure, and sustainable way. It means creating the conditions for agent autonomy to translate not into loss of control, but into real execution capability.

Because when AI starts acting, there is no room left for improvisation.

AI will act. The question is: on what?

Whether we like it or not, AI agents will act.

They will enter processes, platforms, and everyday operational flows. Not as isolated experiments, but as structural components of digital systems.

The real question, therefore, is not if it will happen. It is how.

Which APIs will agents be able to act on?

With what level of control and visibility?

Within which responsibility model?

These are the questions that today separate conscious AI adoption from a frantic race after innovation.

At ApiDays, ApiShare deliberately chose to shift the focus from hype to foundations. Because if it is true that AI represents the engine of change, it is equally true that governed APIs are the road on which that engine can truly run.

And roads, as we know, are not improvised when traffic is already moving. They are designed first.