<!--
  Translated by the docs agent from content-src/es/guides/agentes-ia/01-que-es-un-agente-ia-y-como-se-organizan-sus-instrucciones.md
  (source_hash 16b4d6f2101e). If that Spanish source_hash changes, re-check this translation.
-->

An **AI agent** is who talks with your customers on your behalf over WhatsApp, Telegram and your website
chat: it understands what they ask, replies with the information you gave it and, if needed, hands the
conversation off to someone on your team. Everything that decides how it behaves is configured from the
**AI agents** screen, in the sidebar.

Your instructions don't start from zero. When you signed up your account (**Manual 22 · Account sign-up**)
you chose a **business type**, and with that the Platform already created you a first agent **adapted to
your sector**, with access to everything you'd upload and a web channel ready to embed — that's the
foundation the platform lays down. From there, you add your own: from this screen you adjust its name, its
tone, its instructions (the **system prompt**, section 3), what it replies when it doesn't know something,
when it should escalate a conversation (section 4) and which documents, templates and connections it can
use (section 6). You can create as many agents as your agency needs — one for sales and another for rentals,
for example — and assign each one to the channels you want (section 5).

The screen has a **master-detail** layout: on the left, the list of configured agents, with their name,
model and status (**Active** or **Paused**); on the right, the full editor for whichever agent you've
selected.

![AI Agents screen with the list of configured agents on the left (with their Active/Paused status) and, on the right, one agent's editor with its name, LLM model, tone and system prompt](/captures/agents-editor.png)