Tecnológico de Monterrey: an AI agent that answers from its own information

Where they started

Tecnológico de Monterrey (https://tec.mx) is a university in Mexico with a large amount of information for people approaching it: programs, procedures, requirements and services. It wanted an AI agent that answers questions from its own content and not from generic information.

Atomix built it on its own agent platform, designed so that each organization can have its own agent configured with its own material.

The day-to-day problem

In an institution of this size, the same questions arrive again and again, and the answers are spread across many documents and pages. An agent is only useful in that setting if it answers with what the institution published and doesn't invent what it doesn't know.

What we built

• A custom agent, with a name, tone and instructions specific to the institution.
• Its own knowledge base, built from files, texts and web pages.
• Answers limited to that information: the agent replies with what it was given.
• Website chat, with configurable colors, welcome message and texts.
• Conversation history so the team can review what was asked.

How it works

The institution's material is loaded and the platform processes it so the agent can search it. When a person writes in the chat, the agent consults that material and replies with what it finds, in a tone defined beforehand. Conversations are recorded and the team can adjust the instructions or the content when needed.

Decisions we made

The main one was that the agent answers only from the loaded information. For a university, an invented answer about a procedure or a requirement is worse than no answer. That is why the agent's instructions ask it to rely only on the material it has.

Where it is today

Tecnológico de Monterrey has been using the agent since 2025, for more than a year.