Among generalist or specialist, companies already have their response

He initial astonishment against chatgpt It was not only because of that magical feeling of seeing how character responded to character, pattern inherent to the LLMs that humanized them to some extent. The astonishment was because they knew everything. They explained quantum theory and wrote poetry, summarized novels and armed a business plan in seconds.

They seemed capable of anything, such as the classic first row student who dazzled because he analyzed Blasco Ibáñez with the same precision with which he resolved a differential equation.

The question, sooner or later, always comes: What is the use of?

In it Deloitte technological trends report for 2025 A track appears: many companies that had opted for these generalist models – large, complex, difficult to refine – are beginning to look at smaller and specific options.

Models trained with less data, but much more relevant. Specialists, no todologists. It is no accident: that initial enthusiasm with the Llm It is running with a reality: knowing everything is not always useful. And the world of business does not value wisdom, the margin is valued.

As sometimes it happens, this is a more philosophical change. And he looks a lot like an old debate in companies: that of human specialists vs. generalists.

• The expert who has dedicated his life to a single field that dominates as anyone …
• … Faced with the broad, curious, adaptable profile, with tangential knowledge.

David Epstein explored it well in a book that I loved it‘Amplitude‘. That title made a somewhat uncomfortable thesis fashionable: in a changing world, specialization can become a cage. But the AI, perhaps in a counterfit, is returning luster to the specialist profile.

Because? Because in practice, Generalist models are vague. They try everything but refine a little. An AI that advises doctors, lawyers or engineers cannot be improvised. It needs rigor. Context. Know the terrain. And that does not give it, the approach gives.

There is a slightly thinner reading here. The turn to specialized models allows greater efficiency, but also more control. The big models are in the hands of a few: OpenAi, Google, Anthropic, goal … are closed, opaque, often expensive.

The smallest models can be open, trained at home, adaptable to concrete niches. They look more like tools than oracles.

It also has labor implications:

• If a generalist can do “everything”, it is a diffuse threat.
• If there are many specific ones, they may not come so to replace people, but to expand them.

A doctor with an AI adjusted to his specialty, an architect with an assistant who knows how to read plans, an –ejem– editor with a co -pilot specialized in his sector. It is not the same to compete with a universal AI as collaborating with a refined tool.

And this connects with something more important: A new knowledge economy. For years, we were told that “knowing everything.” Be versatile, navigate between disciplines. Now, companies reward the located, technical, deep knowledge. We already know that IA transforms workbut perhaps also our ideas about knowledge. What is worth, what matters.

And there comes the question. What kind of intelligence do we want to enhance? One who knows a bit of everything and monopolis attention? Or many humble intelligences, distributed, each focused on solving their own problems?

Choosing between generalists or specialists is how we want to live with AI. And what knowledge model we prefer for the coming world.

Outstanding image | Elen Sher and Patrick in Unspash

In Xataka | Deep Research is not just a new AI function. It is the beginning of the end of intellectual work as we know it