Analysis of the article published on SSRN by Luciano Floridi

In 2025, the debate around artificial intelligence seems frozen between two opposing narratives. On one side, an almost messianic promise: that of increasingly intelligent machines, capable of reasoning, deciding, even replacing certain forms of human judgment. On the other, a diffuse anxiety: loss of control, dilution of responsibilities, blind automation. In this hubbub, a recent academic article brings a welcome conceptual clarification. Its thesis is simple, yet fundamental: artificial intelligence is not intelligent. It acts without understanding.

This semantic shift, which may seem abstract, has very concrete implications for entrepreneurs, executives, and decision-makers deploying AI systems in their organizations today.

Moving Beyond the Myth of Artificial Intelligence

Floridi starts from a methodological observation. The term “artificial intelligence” is historically loaded. It suggests cognitive capabilities comparable to those of humans: understanding, intentionality, autonomous reasoning. Yet this analogy is misleading. Current systems, even the most advanced ones, do not think. They do not understand. They have neither consciousness nor their own intention.

What they do, however, they do with remarkable efficiency. They classify, predict, generate, optimize. They interact with their environment. They produce real effects in the economic and social world. For Floridi, the right analytical framework is therefore not that of intelligence, but that of the capacity to act.

AI as Artificial Agency

The article’s central idea rests on an essential conceptual distinction. Floridi contrasts two ways of thinking about AI. The first consists of extending the notion of intelligence to machines. The second, which he defends, consists of recognizing that AI primarily possesses an artificial capacity for action.

This capacity for action refers to the fact of acting in an environment, producing effects according to rules, objectives, and constraints, without it being necessary to associate understanding or intention with it. An organization, a market, or a computer protocol can act without being intelligent.

From this perspective, AI appears as a delegated, framed, and strictly instrumental capacity for action. It executes objectives without its own intention. It optimizes processes without understanding the meaning of what it optimizes.

Why This Distinction Is Crucial for Entrepreneurs

For an entrepreneur, this clarification is not philosophical in an abstract sense. It touches the heart of decision-making. Many organizations today attribute to AI systems an implicit status of quasi-decision-maker. We speak of AI “that decides,” “that recommends,” “that arbitrates.”

According to Floridi, this way of speaking introduces a dangerous bias: that of illusory cognitive delegation. An AI system does not exercise judgment. It implements a programmed, trained, and parameterized capacity for action. The objectives it pursues are never its own. They are always the product of prior human choices: choices of data, models, performance criteria.

For executives, this implies reinforced, not diminished, responsibility.

Governance: Regaining Control Over Delegated Action

One of the article’s most operational contributions concerns governance. If AI acts without understanding, it cannot be held responsible for its effects. Responsibility remains structurally human.

Floridi insists on this point: speaking of AI “autonomy” is often a misuse of language. It is at best a limited functional autonomy, always framed by architectures, objectives, and thresholds defined by humans or organizations.

For a company, this means that any automated decision engages those who designed, validated, and deployed the system. There is no possible moral or legal transfer to the machine.

Avoiding Managerial Anthropomorphism

The article also warns against a frequent drift in corporate discourse: anthropomorphism. Attributing intentions, understanding, or a “vision” to AI leads to a loss of critical vigilance.

Floridi shows that this projection is intellectually comfortable. It allows technical choices to be naturalized and certain decisions to be presented as inevitable. In reality, these decisions are always the product of a design framework.

For entrepreneurs, the question then becomes central: what choices have we encapsulated in our systems? What trade-offs have we frozen into models? And what happens when the context changes?

AI as a Strategic Tool, Not a Cognitive Substitute

Another merit of the text is to reposition AI in its proper strategic place. Floridi in no way minimizes its power. He describes it as extremely effective in certain tasks: massive data analysis, pattern detection, automation of complex processes.

But this power is that of execution, not meaning. For companies, this implies a clear articulation between human decision and automated action. AI can enhance the capacity to act. It cannot alone define what is worth doing.

This distinction is crucial in sensitive domains: recruitment, healthcare, finance, predictive justice, investment strategy. Where normative stakes are high, automated action must remain explicitly subordinate to human governance.

A Timely Reading in the Era of AI Agents

At a time when “AI agents” are becoming an omnipresent keyword, Floridi’s article arrives at the right moment. It helps to overcome a dangerous conceptual blur. Not all systems capable of acting are intelligent. Not all complex behaviors involve understanding.

For entrepreneurs, this analytical framework offers a real competitive advantage: designing more robust, more responsible, better-governed systems. It invites asking the right questions upstream: what capacity for action do we wish to delegate? How far? With what control and accountability mechanisms?

Conclusion: Less Illusion, More Lucidity

Luciano Floridi’s article does not seek to slow innovation. It seeks to make it conceptually rigorous. By rejecting the illusion of autonomous artificial intelligence, it restores the full place of human responsibility.

For entrepreneurs, the message is clear. AI is not an external brain. It is a powerful lever of action, but blind to meaning. The maturity of AI in business will not be measured by its supposed level of intelligence, but by the strategic lucidity with which it is designed, governed, and used.