Red Hat Summit 2026: “Today Artificial Intelligence is slowing down, but in two years it will turn around”

The Brazilian Gilson Magalhães and the Argentine Jorge Payró, top managers of the region of the emblematic open source company, chatted with Clarion about the current state of AI in large companies. Red Hat provides AI solutions to diverse firms: banks, automotive companies, prepaid medicine companies, energy companies, government agencies and even NASA.

-AI models such as Gemini or ChatGPT are already very present in the daily lives of individuals. Why doesn’t the same happen in companies? Such a strong adoption is not seen in companies.

-“For most companies, AI is not yet completely ready. But it will be in two years. Within two years, companies will massively adopt AI. Because today AI gives us a lot of work for human supervision. Above all because of the care that must be taken with What types of data do companies provide to AI?. So today she is not totally ready, we still don’t have confidence in her. And my forecast is that in two years There are going to be AI agents that are going to have the ability to supervise governance and digital sovereignty issues, so that the company has full control of its AI. And only then are companies going to let go,” says Gilson Magalhães, VP and General Manager of Red Hat Latin America.

Jorge Payró, Country Manager of Argentina & Regional Manager of SOLA-East at Red Hat, agrees but does not allow himself to be so forceful with the deadline. He doesn’t see a D-day.

-“I think that in certain points there is still a lot to be done. For example that ability to create that humans have today and Generative AI does not have, which is always based on prior knowledge. Today you can get up (he addresses this chronicler) and say ‘I’m going to make such an article, I’m going to create this in such and such a way’ and it is yours, original, your own, and perhaps it takes previous concepts but is unique from the point of view of creation. That point has not yet reached Generative AI. And it will take more years until it can play a decisive role in everyday life. I believe that today there are still many limitations and why not say it, there are also interests around what is happening and what is happening. We have to see how it develops. What I do believe is that evolution is at accelerated pace,” he says.

-“Of course,” Magalhães agrees, “AI is more than two years away from creation. In a business discussion AI is going to create new offers. They may show new businesses that humans did not see. It will show opportunities.

-What are companies failing?

-“A lot of data was given to AI but without a clear objective, without a direction. Many companies got into AI because they had to have AI, out of fashion, but they don’t really know why. So you have to have a goal. And when you know what your objective is, you can give the AI ​​the data it needs for that objective and you will be able to control it better. That was not done from the start. That’s why AI is slowed down today. But In two years we will see the change”, says Magalhães.

-Why is AI mature in some areas, for example banking, and not in others?

-“Because there they have a specific objective. For example, they want to make a bot for customer service. And that is done perfectly. It provides information, etc. There is a clear objective there,” explains Magalhães.

“I believe that there is always early adopters in the industry. There are some industries that, due to current competition, consider that by introducing certain technology tools, they can gain an advantage over the competition. On that path the financial industry (of which there are already many cases in our region), the telecommunications industry. And then, depending perhaps a little on geography, energy, health and aerospace companies. In medicine there is already collaborative agents of professional doctorswhich they can do as a prior diagnosis with an image. And then it is obviously supervised by a human, the professional doctor who says, ah, yes this is indeed the case,” adds Payró.

-What could AI agents contribute?

-“What we cannot eliminate yet is the possibility of hallucinations in AI. That will be present. So, how do I work to correct those hallucinations and separate, let’s say, the information that is correct, valid, useful from other information that is not, that is inferred through data that has not been analyzed in the correct way or in the correct criteria.

I believe that when they are trained, the supervising AI agents will be able to detect these hallucinations. Work is already underway on agents that can detect hallucinations. The progress is very fast. A company that is a client of ours last year had 30% of the code they make done by AI. This year it is already 60%,” concludes Magalhães.