The agents are the future that AI companies promise and urgently need.

Automation of tasks has been a constant throughout human history, and today, artificial intelligence companies are channeling this need for efficiency under the concept of "agents." AI agents are autonomous programs that can perform tasks, make decisions, and interact with various environments with minimal human intervention. This trend has captured the attention of the major players in the field of artificial intelligence.

For example, Microsoft has developed "Copilots," a tool aimed at facilitating the automation of processes such as customer service and administrative tasks. In another notable move, Google Cloud CEO Thomas Kurian has proposed six AI-based productivity agents. Google DeepMind, on its part, has recruited Sora, a co-leader at OpenAI in video product development, with the goal of creating simulations to train AI agents. Anthropic has also made strides by allowing its users to create an "AI assistant" through its chatbot Claude.

While many autonomous systems already exist, like online customer service bots or voice assistants, AI companies want to emphasize that their agents are different. Unlike bots that follow basic procedures, agents are designed to interact, learn from experiences, and make decisions more independently. They are expected to dynamically manage a variety of tasks, from shopping and travel bookings to scheduling meetings, adapting to unforeseen situations.

Artificial intelligence companies see these agents as an opportunity to monetize sophisticated and costly AI models. There is heavy investment in startups that promise to revolutionize technology interaction. The vision is for agents to increase efficiency in businesses by taking on tasks that range from customer service to data analysis. For individuals, the idea is to usher in a new era of productivity by automating routine tasks, thereby allowing them to focus on more creative and strategic work.

OpenAI CEO Sam Altman has described an ideal application of AI as a "super-competent colleague" who has comprehensive knowledge of your life, capable of performing basic tasks quickly and tackling more complex ones, though it may require asking additional questions. Although companies have sought this advancement since the 1970s, they claim to be closer than ever to achieving it.

In a recent demonstration by OpenAI, Romain Huet, the developer experience lead, showed how an agent could place an order for chocolate-covered strawberries, even managing conversations in Japanese to book a hotel. However, the presentation raised concerns about the potential misuse of these tools for unsolicited calls and the lack of transparency, as the previous version did not identify itself as an AI system.

Moreover, the demonstration encountered failures, highlighting that agents still face significant challenges in complex workflows. Despite their potential, their performance on everyday tasks is not significantly superior to that of current bots or assistants. Companies are working to enhance their capabilities through reinforcement learning, but the high costs of operating at scale present a greater obstacle.

Although the idea of agents has gained popularity due to market pressures, their real effectiveness remains a topic of debate. There are questions about their ability to handle sensitive tasks such as legal or tax issues in a context where AI "hallucinations" remain an unresolved problem. It is crucial to remember that AI assistants should be considered powerful but imperfect tools, especially for low-risk tasks.

Interest in these systems is growing, and according to the view of several industry experts, such systems are expected to become commonplace by 2025, offering the possibility of spending more time on activities that truly matter instead of getting tied to technological devices.