It's Not Just GenAI: Autonomous Agents Are Coming to Life

In 1995, in his first book, “The Road Ahead,” Bill Gates predicted that the technological race would evolve to a level where autonomous agents would be present across every device — in search engines, e-commerce platforms, word processors, spreadsheets, apps, and the code of multiple digital spaces — and would assist humans in solving virtually all their tasks.

Today, 30 years later, companies are seeking to transform autonomous agents into a reality, and in fact, they are far more used than most people imagine. However, it's still difficult to quantify the real impact they can have on certain industries and their return on investment.

What Are AI Agents?

AI Agents are what Gates once described as autonomous agents; they are the natural evolution of generative artificial intelligence. Although they are developed with machine learning, they are more advanced assistants, capable of learning from interactions and managing tasks more independently.

In recent years, many companies have improved customer service thanks to chatbots that answer tasks based on the information their language model allows them to process. When a request is received, they interact and then create a support ticket to be resolved by a human.

Do they make the task easier? Of course they do. But AI Agents will also be able to execute it. In other words, if the customer requests a refund, they will be able to interpret the request, determine whether a return is appropriate, and if so, process it. Agents learn from interactions, adapt to new scenarios without human intervention, and are capable of solving complex tasks.

The Paradigm Shift for all Industries Is Beginning

By 2028, one-third of generative AI interactions will involve autonomous agents, according to a Gartner estimate. In a recent MIT Technology Review survey of 300 business leaders, half of them consider that the main benefit from autonomous agents is that they help improve efficiency and save costs, while 20% believe that their main benefit is helping improvements or going faster than competitors.

In some industries, the productivity increase is estimated to be as high as 30%, which represents an economic injection of US$15 billion into Mexico's overall GDP. While adoption is increasing, there are significant opportunities.

The ways to understand AI Agents’ efficiency vary depending on the industry in which they are used. In the world of information technology, for example, they provide efficiency by understanding the number of code lines written daily, the total cycle time for a change, the review time, the reduction in the use of obsolete software versions, the average repair time, or the time spent searching for information in coding tasks. In short, they allow for better results with the same investment. This example can be applied to any scenario where an employee is performing a task using a computer; for example, order receiving centers, claims centers, production scheduling offices, quality control spaces — the scenarios are endless.

In all cases, AI Agents act as a chain of workers who, simultaneously, while one collects data, another can interpret it, and a third can present multiple forecast scenarios for a given situation. This ability to interpret the problem during the process and subdivide tasks to solve them all simultaneously represents an unprecedented advance in the history of technology.

While all this is happening behind the scenes, there are already companies taking advantage of them. We can think of ride-hailing apps like Uber or DiDi, which dynamically set fares without human intervention in real time and simply by interpreting supply and demand. It's not far-fetched to imagine AI Agents deciding on parts assembly in a factory, adapting educational content to students' needs, providing a doctor with a patient's preliminary diagnosis, or developing exclusive promotions for certain customers without any employee having to design them.

But What Are AI Agents’ Challenges and Responsibilities?

With the fast advancement of AI, questions arise about its accountability, safety, and fairness. AI agents are tools, not responsible by themselves. If a lawyer uses AI and obtains erroneous information, the liability remains with the lawyer, not the system. Suppliers must implement controls to prevent errors, but user trust is key for adoption. Although the law on accountability is still unclear, it is essential that companies manage these systems ethically.

Regarding security, AI agents must be protected within a system that adequately controls them.

Bias in AI is a severe problem. As technology becomes more complex, it is more difficult to prevent systems from discriminating. A clear example is facial recognition, where a system can reject access to people who do not conform to a specific appearance standard. This can lead to unfair situations, where a system denies access or services to someone without a valid reason. To avoid these problems, it is essential to offer clear explanations and alternatives so that affected individuals have the opportunity to correct errors. AI agents are no longer just tools that facilitate tasks, they are being integrated as part of a team, adapting to the needs of each company. In sectors such as healthcare and finance, they are accelerating decision-making and improving efficiency. In healthcare, they help process scientific data and optimize treatments. In finance, they manage risks and improve users' financial health. But their implementation must be done ethically and responsibly to prevent errors and discrimination from recurring.

In 2025, autonomous agents or AI Agents are not yet as real as Bill Gates predicted, but they are on the fast track to becoming so. The challenge will be for companies to know where and in what sequence to implement them. And, of course, to measure the benefits of doing so.