Artificial intelligence: what AI agents are and how they will impact the way we work

In the last three years, AI has evolved rapidly, generating changes in the way we work. What will be the impact of AI agents now?

First we must define what an agent is.

Diego Castronuovo, AI leader at the software company N5, indicated in a report by El Comercio that the AI ​​agent is one of the key factors for this year, characterized because “they are dedicated exclusively to taking a task, analyzing it, dividing it into steps and executing it”. On the other hand, AI specialist Wester Zela explained that it is a new technique to make AI have fewer hallucinations (errors), better responses.

Specifically, an AI agent can be said to be a software program that can interact with its environment, collect data and use it to autonomously perform defined tasks in order to meet predetermined objectives.

“This means that the agent can carry out its functions without the need for constant human intervention. The agent makes its own decisions based on the information it perceives from the environment, the predetermined objectives and the rules or models with which it has been designed.indicated Jesús Bellido, dean of the Faculty of Computing at UTEC.

One of the most current examples can be seen in the operation of so-called autonomous cars, or smart cars, which can decide to brake, turn or accelerate depending on the traffic situation.

How do they work?

An AI agent can encompass both software and hardware. This shows the importance of architecture in the development of an agent, especially for the task it must fulfill.

As it is software, it works using a lot of data to analyze it, but when combined with a physical structure it can also consist of actuators, sensors, motors and robotic arms, with which it can obtain information about its environment.

Another feature is autonomy. This means that the agent can work independently, although it needs to be programmed (human action). “When designing the agent function, developers take into account the type of information, AI capabilities, knowledge base, feedback mechanism and other necessary technologies”Bellido tells this newspaper.

As with the well-known chatbots or generative AI tools, these AI agents also need to be trained to simplify different tasks, they can be differentiated, structured and divided to achieve a better result. Some examples are virtual assistants, chatbots, recommendation systems, autonomous vehicles, etc.

“A recommendation system might not require sensors to make decisions based on the consumer’s history. On the other hand, an autonomous vehicle uses between 20 and 40 sensors, combining different technologies to ensure accurate and reliable perception of the environment.”says the UTEC specialist.

Impact on work

There are many ways to apply AI agents. The result is to improve a task, or production in a company, but does this mean a risk for workers?

On the one hand, the AI ​​agent can be an ally by reducing hallucinations, or errors that some artificial intelligence tools have today, many of them due to obtaining outdated or biased information from the Internet. Additionally, it can function as an assistant by automatically replying to emails, researching, reporting, or other programmable tasks.

It also involves adaptation. “An AI agent can both complement and compete with human work. While it can replace certain tasks, it also opens up new possibilities by allowing people to focus on higher value-added activities. The challenge is to prepare the workforce to adapt to these changes through training and continuing education”says Bellido.

More than a threat, much will depend on how people adapt and coexist with artificial intelligence.