Originally published on Administradores.
Cycles of enthusiasm followed by frustration are part of a recurring pattern in the technology market. Over the past two decades, we have watched different promises gain traction and lose relevance, as in the early 2000s, when platforms such as Second Life (launched in 2003) mobilized companies and users around the idea of persistent virtual environments. There was investment, experimentation, and even tests of business models.
Years later, between 2021 and 2023, the metaverse returned to the center of the agenda. Large companies announced billion-dollar investments, repositioned strategies, and tried to accelerate the building of that new digital environment. But none of those waves took hold as a structural foundation of the market, even though we cannot deny they mattered as experimentation. The fact is they did not consistently change the way people work, produce, or engage with one another economically.
Artificial intelligence is in another category.
The central difference is that it does not depend on a specific environment to exist. It integrates into what already exists and changes how those structures work. That is already visible in the way work itself is done: the change is not only about gaining efficiency, but about changing the role of each step. The numbers bear this out: a McKinsey survey of nearly 2,000 organizations across 105 countries shows that 78% already use AI in at least one business function, up from 55% in 2023.
In technology, for example, code is no longer the end goal. It stops being the limitation that separates those who can build from those who cannot. People who understand product, flow, and experience start structuring solutions even without deep programming skills, because execution can be assisted.
The same happens with analysis. Activities that once depended on weeks or months of collecting, organizing, and reading data, often run by specialized consultancies or dedicated teams, can now be handled with much less friction, without the same dependence on manual processes.
In content production and market intelligence, the logic also changes. Access to data, the ability to synthesize, and the production of materials no longer require the same operating structure that existed before.
The effect of that is not only speed. It is less dependence on manual execution as the main asset of the work and, of course, more weight on whoever understands the problem, the context, and the decision that needs to be made — radically affecting efficiency and who can do what.
Professionals who understand the problem start operating with more autonomy, even without deep technical skills, and execution stops being the main bottleneck, shifting value toward understanding. At the same time, the transformation is starting to reshape how systems and individuals connect.
Historically, technological evolution has been marked by changes in communication protocols. In the early days of system integration, the dominant standard was SOAP — robust, but complex and inflexible. Over time, REST took hold as a simpler, more efficient, and widely adopted alternative. That shift was not only technical. It made integrations scalable and let companies connect more quickly. And now a new layer is beginning to emerge.
The trend is toward agent-based communication models, such as what is taking shape around protocols like MCP (Model Context Protocol). In that scenario, it is not only systems that connect; agents represent individuals and organizations. Do you see how that changes the logic of interaction?
It is not theory: MCP, launched by Anthropic at the end of 2024, was adopted within a few months by OpenAI, Google, Microsoft, and Salesforce, moved under the neutral governance of the Linux Foundation, and jumped from about 100,000 to 97 million monthly SDK downloads in 18 months. At Celcoin, we already operate infrastructure on that model.
Today, one company integrates with another. In the near future, personal agents may interact directly with services, negotiate, execute tasks, and make decisions based on context. The interface stops being only an application and becomes an active intermediary — and that is what makes this kind of change deeper than any previous trend, because it alters the operating structure, not only the experience layer. And that has already left the drawing board: 23% of organizations have already scaled some agentic AI system and another 39% have started experimenting, or 62% in total.
Another relevant point is speed. In the first wave of the internet, between the mid-1990s and the early 2000s, learning was slow for lack of access to information. Professionals learned by trial and error, without clear references. Today the picture is the opposite.
There is an excess of information, accessible tools, and AI itself as support for learning. That drastically cuts the time needed to acquire new skills, which directly compresses transformation cycles.
Changes that once took years now happen in months. But for a long time, quality was tied to time, and well-executed projects demanded long cycles of development and refinement.
With AI, that relationship changes, because the expectation becomes speed with quality.
Companies are already reviewing how they produce content, develop systems, and structure analyses, not as an experiment, but as an operational adjustment. That repositioning tends to redesign roles, as some activities no longer depend on manual execution and start being absorbed by new ways of working.
As in other moments in history, the introduction of a new technology tends to redesign the labor market. Some roles stop making sense, others are transformed, and new ones emerge from that reorganization. That movement is not new — it is what Schumpeter called creative destruction: the technology that creates value is the same one that makes earlier forms of work obsolete. What sets the current moment apart is the speed and, above all, the breadth with which this happens.
That is because artificial intelligence does not act on an isolated layer of the economy. It is not restricted to a specific sector or a narrowly defined function. On the contrary, it crosses different industries and simultaneously impacts operational, analytical, and creative activities. That widens the reach of the change and leaves less room for comfort zones.
But it is worth recognizing a legitimate objection. Back in the 1980s, economist Robert Solow joked that computers were showing up everywhere except in the productivity statistics. That was the so-called productivity paradox; and something similar is visible today. Despite massive adoption, most companies remain stuck in pilots that never scale, with recent studies even finding that professionals who felt faster with AI were, in practice, slower. That finding, however, does not weaken the argument. It reinforces it. The gain does not come from bolting the tool onto the old process. It comes from reorganizing the entire operation around it. And that is exactly why the discussion shifts from the technology to the capacity to adapt.
For that reason, the discussion is no longer about whether there is an impact. On AI, that point is already settled. The central question becomes how well professionals and companies can adapt in a scenario where the rules of production, decision-making, and delivery are being rewritten in real time.
Resisting change does not keep it from happening; it only postpones the reckoning. And, historically, late adaptation usually carries a high cost, precisely because it demands a more abrupt correction in an environment that has already moved on.
Unlike other recent waves, artificial intelligence does not present itself as a future bet or as a trend still in a validation phase. It is already embedded in operations, in decisions, and in the way value is generated. It does not depend on creating a new environment or on adopting a new logic.
What is underway is a structural change: artificial intelligence is redefining how the existing environment works. Earlier waves passed because they were environments waiting for adoption. AI stays because it is already the layer on which everything operates.