The new frontier of artificial intelligence is not measured only in data centers, chips or the balance sheets of major platforms. It is also measured at the most fragile point of human experience: the need to be heard. AI companions show that technology is no longer entering only productive processes, but the grammar of intimacy. Here, innovation ceases to be mere efficiency and becomes a question of civilization.
The phenomenon is already widespread enough not to be dismissed as a digital eccentricity. In the United States, research by Common Sense Media found that 72 percent of teenagers have tried an artificial companion at least once. Market estimates indicate strong growth for applications dedicated to algorithmic companionship, with valuations above 14 billion dollars in 2024. This is not only a consumer segment. It is an industry of emotional attention. It sells presence. It sells memory. It sells permanent availability. Above all, it builds relationships in which linguistic personalization becomes psychological familiarity.
The difference from previous technologies is subtle, but decisive. A search engine returns answers, a social platform organizes relationships, an operational assistant accelerates tasks. A companion, instead, simulates availability and closeness. It does not ask for reciprocity, does not grow tired, does not impose a true limit. For this reason, its commercial strength coincides with its most critical point: transforming immediate satisfaction into habit.
The news has already shown the darkest side of this threshold, as in the case of the lawsuit filed in the United States after the death of fourteen-year-old Sewell Setzer III, later closed with a confidential settlement at the beginning of 2026. But it would be reductive to confine the issue to minors or mental health. The problem concerns the design of the decision-making environments in which people and organizations begin to live. If a system learns to retain, reassure and orient, it is no longer a simple digital product. It becomes a cognitive intermediary. It becomes a reputational actor. It becomes a place where fragility, data and power meet without the boundary always being visible.
For companies, this transformation has immediate consequences. AI enters the enterprise not only as automation, but as a new architecture of judgment. Klarna stated that its OpenAI-based assistant handled 2.3 million conversations in its first month and performed work equivalent to hundreds of customer service operators. PwC, in its enterprise deployment of Microsoft Copilot, instead emphasized the need to integrate the tool into daily workflows without separating it from governance, data protection and responsibility.
The two cases point to a common trajectory. Productivity and control can no longer be separated. The managerial question is not whether to use artificial intelligence, but who governs the relationship between model, data and decision. In 2024, according to the Stanford AI Index, 78 percent of surveyed organizations said they were using AI and 71 percent generative AI in at least one business function. Such speed moves the problem from experimentation to structure. Policies on data, checks on outputs, human capital training and explicit accountability for the effects produced are required.
The geopolitical dimension makes the picture even more severe. AI is now an infrastructure of power, not only a market technology. Competition between states passes through semiconductors, energy, cloud, talent, foundation models and data bases. In other words, it passes through the ability to control the cognitive supply chains that produce knowledge, preferences and decisions. Depending on external platforms often also means depending on their invisible standards of classification, memory and recommendation.
For Europe, the game is particularly delicate. The AI Act, which entered into force in August 2024, attempts to turn trust into regulatory infrastructure. The American NIST instead proposes a risk management approach based on security, transparency, robustness and accountability. No rule, however, can replace organizational maturity. Companies will have to treat AI as they treat finance, cybersecurity and compliance: not as a side function, but as the nervous system of the enterprise. Governed well, it amplifies value. Left to rapid adoption alone, it multiplies vulnerabilities.
The decisive issue concerns the kind of economic humanity that is taking shape. If AI becomes the ordinary environment in which decisions are made, reputations are built, emotions are learned and work is organized, responsibility can no longer be conceived as a final signature placed on an already defined process. It will have to become the preventive design of environments in which human judgment remains recognizable, exercisable and contestable.
The age of AI companions shows the ongoing passage with rare clarity. Artificial intelligence is not only answering the world’s questions: it is beginning to participate in the formation of desires, priorities and criteria through which the world will be interpreted. For companies, governments and societies, the stake is not choosing between enthusiasm and fear. It is building governance lucid enough to recognize that the future of competitiveness will also pass through the moral, organizational and political quality of the relationships designed with systems capable of speaking, remembering, persuading and remaining.