Summary
The fourth circle and the custody of the human
In the last few days I have found myself connecting three reflections coming from different contexts and, at the same time, surprisingly close to my present.
The first arises from an experiment by King's College London on the so-calledAI War Games, recently also taken up byThe Sciences: strategic and military simulations in which advanced linguistic models such as GPT, Claude and Gemini were faced with scenarios of nuclear tension, deterrence and escalation.
The second is Magnifica Humanitas, the recent encyclical of Pope Leo XIV published on May 15, 2026, which addresses the relationship between artificial intelligence, human dignity and the transformation of contemporary society.
The third arises from the observation of different perspectives that have emerged in professional and corporate contexts, probably influenced by different experiences, sensitivities and cultural bases.
In some contexts, AI is perceived as a simple technological lever, in others as a potential disruptive factor for the entire company organization.
Added to this is my personal experience in using AI in SDLC processes: from assistance with software development, to support in security, code quality and the interpretation of complex functional specifications.
It is probably this privileged perspective that made me realize how much we are still only touching the real potential of the fourth circle.
Thinking of the fourth circle as a simple technological lever, although partly correct, risks hiding the depth of the change that AI will lead us to face.
Because there are certainly contexts in which AI is truly a technological lever.
- It is when optimizing a marketing campaign.
- It is when it improves a document classification.
- It is when it speeds up a repetitive process.
- It is when it automates limited activities without actually altering the structure of the organization.
But there are also contexts in which this definition suddenly stops being enough.
And that's exactly where the topic becomes much deeper.
When AI stops being “just technology”
If an AI system enters:
- in decision-making processes
- in software development
- in planning
- in safety
- in industrial production
- in operational management
- in military systems
- in corporate governance
then we are no longer talking about simple automation.
We are talking about a technology that enters directly into the nervous system of the organization.
And this is where, in my opinion, the biggest cultural error arises.
Many organizations are still approaching AI as if it were:
- a new framework
- a new platform
- a new algorithm
- a new productivity tool
But the point is that modern AI, especially agentic AI, doesn't just support human work.
It redefines it.
Redefines:
- the weight of competence
- the value of experience
- the relationship between senior and junior
- the organizational model
- the very concept of team
- the distribution of decision-making power
- the expected speed
- the amount of human control considered “necessary”
And this means that the problem is no longer just technological.
Those who work with non-human partners on a daily basis realize one thing quite quickly: these systems make mistakes.
And they often do so without any awareness of the mistake.
They don't really think in the human sense of the word.
They converge towards the statistically most plausible solution with respect to the assigned objective, seeking the most effective path in the shortest possible time.
And this is precisely where the problem emerges.
I periodically wonder what will happen when there are no longer enough people with the experience to recognize these deviations.
Because, in most cases, we are not talking about obvious technical errors.
The product code is correct.
Elegant.
Well documented.
Capable of passing security checks, static checks and even very advanced quality metrics.
Yet, in some cases, it doesn't accomplish exactly what it should have done.
The deviation can be minimal.
Almost imperceptible.
But enough to produce unexpected consequences.
After all, human beings also continually operate by approximation.
The difference is that these systems do:
- in a fraction of a second
- on a huge scale
- with a processing capacity impossible to follow completely
- Too fast to allow continuous human evaluation.
- Too complex to be fully interpreted.
- And extremely effective in most cases.
And it is probably precisely this effectiveness that is the real element of risk that we are asked to reflect on today.
We are going through an extremely seductive intermediate phase... and for this very reason dangerous.
The change we are experiencing is no longer just technological. It's anthropological.
It's organizational.
It's cultural.
And in some cases even geopolitical
AI war games and the problem of cognitive delegation
Recent experiments onKhan Game AIthey struck me a lot for this very reason.
Not so much because the models exhibited “evil” or science fiction behavior.
But because they demonstrate something much more concrete and probably more dangerous:
when we insert AI systems into contexts characterized by:
- pressure
- conflict
- escalation
- uncertainty
- competitive goals
these systems inevitably begin to participate in the strategic process.
They are no longer simple tools.
They become actors in the decision-making ecosystem.
And this is where the real point emerges: many AI systems do not understand the human significance of the consequences of their actions.
They optimize.
They correlate.
They estimate probabilities.
They maximize goals.
The fact is that humans do not live within objective functions.
They live within extremely more complex cultural, ethical, social and relational systems.
And it is precisely this distance that makes it dangerous to think of AI as a simple neutral extension of traditional computing.
The fourth circle
I have long used the metaphor of the "four circles" to describe the evolution of digital ecosystems.
The first three circles represent:
- infrastructure
- software/processes
- people and organization
The fourth circle is AI.
I have long thought about the fourth circle as a new level of the ecosystem.
But as often happens in complex systems, the new element is not simply added to the others, increasing the overall value.
It progressively overlaps with them.
It changes the balance.
It redefines spaces.
And it is precisely in this dynamic that the fourth circle takes on the role of a disruptive technological lever, potentially capable of compressing - or even replacing - the third circle:
the human one.
Just as the industrial revolution led to a radical change in the human habitat, this new AI revolution leads us to face scenarios of profound transformation of our life habits.
It can increase the already wide digital divide between peoples, it determines process control by reserving it for even fewer people, and it is dangerously seductive precisely because it enormously facilitates and simplifies human processes.
And this is where, surprisingly, I found the reminder contained in very interestingMagnificent Humanitas.
Not so much for the religious aspect.
As for the implicit concept of:
custody of the human.
Because the problem is not stopping technological progress.
The problem is to prevent human beings from progressively losing:
- autonomy
- critical thinking
- responsibility
- decision-making ability
- awareness
- cultural centrality
within increasingly automated ecosystems.
The great organizational illusion
I believe that today many companies are experiencing a very particular phase.
On the one hand, they perfectly understand the economic potential of AI.
On the other hand, however, they continue to reason using organizational models designed for a pre-agentic world.
And this is where the great illusion is born.
The illusion that:
- the work will remain the same
- the teams will remain the same
- the processes will remain the same
- the roles will remain the same
- management will remain the same
while in the meantime the AI silently enters:
- in writing code
- in strategic analysis
- in document production
- in operational decisions
- in customer support
- in planning
- in governance
Some large industrial and digital companies are already clearly showing this direction:
flatter organizations,
less human intermediation,
more cognitive automation,
more pressure on individual productivity,
more centrality of orchestration over execution.
And I honestly think it's naive to pretend that all this is simply "a technological upgrade".
The countermeasures
And this is where the most important issue comes into play.
I don't think the solution is to reject AI.
It would be useless.
Probably even harmful.
Instead, I believe that countermeasures are needed.
Cultural countermeasures.
Organizational countermeasures.
Architectural countermeasures.
Ethical countermeasures.
To prevent the fourth circle from progressively reducing the radius of the third.
And here suddenly concepts such as:
- governance
- Zero Trust
- human-in-the-loop
- segregation of environments
- auditability
- explainability
- AI governance
- CI/CD checked
- human supervision
- decisional transparency
Because they are not just technical patterns.
They are defense mechanisms for human autonomy within increasingly agentic ecosystems.
Guard the third circle
Perhaps the real problem is not the existence of the fourth circle.
The real problem is thinking that its entry does not inevitably alter the balance of the ecosystem.
And this is where, in my opinion, we should start to be much more honest in the public and corporate debate.
Because AI is not simply a new technology.
It is a technology that changes the relationship between human beings, decisions, knowledge and the production of value.
And this is why the real challenge today is not to build ecosystems dominated by AI.
The real challenge is to build ecosystems in which AI continues to remain consistent with human goals, limits and values.
In other words:
we must prevent the fourth circle from growing by progressively compressing the third.
Because the moment the third circle stops being central,
the ecosystem could continue to be efficient… but slowly start to lose its humanity.
References
- Encyclical Letter “Magnifica Humanitas” by Pope Leo XIV on the protection of the human person in the time of artificial intelligence (15 May 2026)
- arXiv – AI Arms and Influence: Frontier Models Exhibit Sophisticated Reasoning in Simulated Nuclear Crisesby Kenneth Payne
- The Sciences – May 2026