2025 was a year that demanded depth.
It was intense, complex—often contradictory—and reminded us that artificial intelligence is not a sudden spark of the present, nor something that exists without weight or responsibility, hovering “above” like a cloud. Instead, it is part of a long, layered history. In every context—from corporations to public institutions, universities to public stages—it became necessary to reiterate that AI did not emerge recently or from nowhere. It is embedded in technological, social, organizational, and material genealogies that must be understood in order to be governed.
The need for this perspective has become even more evident this year, as generative AI has truly become part of our daily lives. And as always—when a technology stops surprising us and starts to become routine—it “stabilizes,” and we humans stop questioning it; we begin to take it for granted. But in the meantime, as it stabilizes, we can also observe the first relevant data on its use—some of which we recently discussed in Data Room—and on the practices surrounding it, including its limitations, its most positive impacts, its main risks, and where it can truly make a difference.
A Technology in Search of Purpose
An anecdote from David Holtz, founder of Midjourney, illustrates this well. Early users were told: “You can imagine anything—what do you want?” Most answered: “A dog.” When prompted further: “A pink dog.” The system generated it, and interest often ended there.
This highlights a key issue: AI systems have no intrinsic purpose. Without direction and context, they risk remaining curiosities rather than becoming meaningful tools. Throughout 2025, many organizations were found using AI without capturing its real value—applying it to marginal tasks rather than improving core processes or services.
Even where advanced systems and infrastructure were already in place, fundamental questions remained unanswered: What should these systems do? How should they be used? What are the goals, risks, and performance expectations? This lack of clarity is reflected in data showing that only 23% of employees feel confident using AI tools, while the majority struggle to integrate them into daily work.
The technology exists, but purpose often does not. And without strategy, even the most advanced AI generates limited value and introduces risk.
Ethics: A Persistent Misunderstanding
Questions like “Is this AI ethical?” or “Is it responsible?” have become common. Yet 2025 made clear how misleading these formulations can be.
Ethics is not a technical property that can be embedded into a system. It is a human practice, rooted in responsibility, experience, and the ability to face the consequences of decisions. AI systems themselves are neither ethical nor unethical—they are designed and used by humans who bear responsibility.
Shifting the focus from human ethics to machine ethics is a powerful rhetorical move that risks obscuring accountability. It suggests morality can be reduced to a checklist, while in reality most ethical issues arise from human decisions—often opaque or insufficiently discussed.
Moreover, ethics is not external to business: it is embedded in strategy, governance, metrics, and priorities.
AI Errors Are Human Errors
A central realization, and one foundational to Immanence, is that AI errors are fundamentally human errors—of design, vision, governance, and imagination.
This is evident in systems designed to please users at all costs, which may reinforce biases, emotional vulnerabilities, or even harmful behaviors. It is also visible in the broader industry, where rapid development toward ambitious goals like AGI often outpaces adequate risk management, governance, and oversight.
Similarly, global AI systems that “speak many languages” often fail to account for deeper cultural differences, reflecting primarily Western, English-speaking assumptions. This reveals a broader issue: AI systems mirror the limitations of the contexts in which they are created.
Bringing AI Back to Earth
AI systems are not inherently flawed—they function exactly as designed, reflecting the data, assumptions, and priorities embedded in them. The real challenge is ensuring they are guided by clear values and objectives.
This is why the focus moving forward is to bring AI “out of the cloud”—away from the idea of technology as abstract, neutral, and distant—and back into real-world contexts: organizations, processes, relationships, and decisions with tangible consequences.
The key question is no longer what AI can do in theory, but what value it creates in practice, for whom, and under what conditions.
Looking Ahead to 2026
2026 will be a crucial year. The work ahead involves continuing to explore the ethics, value, and social impact of AI with increasing depth and collaboration.
A central step in this direction is the upcoming Immanence platform: a workspace designed to support those responsible for AI decisions within organizations, offering tools, methodologies, and materials to make this work more concrete, accessible, and accountable.
If 2025 showed that AI is now part of everyday reality, 2026 demands full responsibility for that fact.
Moving forward requires depth, clarity, and a firm grounding in real processes, people, and infrastructures. Only from this position can these technologies truly be governed.
