You have probably heard of AI agents, right? If this term sounds strange to you, perhaps you have already interacted with these systems in their simplest form: chatbots.
However, the concept of agents goes beyond simple chatbots. They are powerful artificial intelligence tools that are likely to impact the future of how we are currently working in companies.
Trying to simplify the concept of AI agents in very few words: They are artificial intelligence architectures “empowered” with tools.
These programs are capable of observing the environment they are in, interacting with the user via LLMs (for example, ChatGPT), processing information, and achieving specific goals using their tools (for example, analyzing and creating documents, analyzing databases, reading and creating emails, making phone calls, creating presentations, etc.). They are also capable of remembering previous iterations (memory) or retrieving specific information (RAG) to provide more contextualized responses.
Let's go to a practical example of an agent that can assist companies in their supply chain (an area I have worked in throughout my career). It is known that many companies waste valuable time with “firefighting,” that is, analyzing and solving unexpected problems that arise daily in the supply plan.
Now imagine an agent capable of analyzing the bases of your ERP, identifying and resolving problems proactively. Watch the video below:
It seems simple, but what this agent is doing is: receiving user inputs in natural language that is understandable to humans, using SQL commands to analyze the MRP database and purchase orders, and returning the information to the user. The system also understands terms used by the company: in this example, the system understands that at-risk material is material that will have availability issues for the customer in the future (stockout).
See the SQL commands that a user should use in the tables to perform the same analysis:
Interesting, isn't it? Through the RAG technique (similar to a procedure), the system understands the database, the terms used by the company, and executes activities in the data tables.
But that's not all. With the right tools, agents can interact with each other and perform different tasks. In this example, imagine other agents sending emails to suppliers requesting expedited deliveries, or even requesting quick loads from transport companies. Other agents, for example, creating presentations based on the outputs of other agents. The possibilities are endless.
Finally, AI agents are shaping the future of work, optimizing processes and driving efficiency in companies. But as they evolve, new challenges arise. How do you envision the adoption of these agents in your company? What opportunities and concerns do you see on the horizon?