Introduction to Agents

Agents are the heart of Platica. They are not your typical "chatbots" with pre-recorded answers; they are new members of your team capable of understanding natural language, consulting your company's documents, performing tasks, and deciding whether to resolve a question or escalate the issue to a human.

The best part: you don't need to know how to code to create one. You can build it through conversation.


How to access it

To get to the command center for your AIs:

  1. Go to the left-hand navigation bar.
  2. Click on the Robot icon.
  3. You will see the list of agents active in your workspace.
  4. Click "Create Agent" to start a new one, or pick an existing one to edit it.

The Anatomy of an Agent

To configure a successful agent, it helps to understand how it's built. As we saw in the video, we can break an agent down into three vital components:

1. Instructions

This is the Prompt. Here you define the personality (formal or friendly), the goal (sell, support, schedule), and the golden rules it should never break.

2. Knowledge

This is the information the agent can consult. Unlike other systems, the agent does not "memorize" all the text; instead, it uses this library as a reference whenever the user asks a specific question.

3. Subagents

These are specialized agents you can create for specific tasks (for example, a billing subagent or a technical support one). The main agent can delegate questions to these subagents whenever necessary.

4. Assistance

Defines the operational limits: What happens if it doesn't know the answer? When should it hand the conversation over to a human?

5. Follow-up

Configures the agent's behavior to proactively reach out to the customer if they stop responding after a defined period of time.

6. Channels

Connects the agent to the communication channels (WhatsApp, Instagram, Web Chat) so it can start interacting with your customers live.

7. Settings

Edit the agent's basic information, such as its name, description, language, and time zone.


If this is your first time creating an agent, we suggest following this logical order to make sure it works properly:

  1. Chat with the Assistant Use the assisted creation tool. Tell the AI about your business and let it draft the first version of the prompt for you.

  2. Load your Knowledge Upload your PDFs, manuals, or web links. Make sure to give each one a clear ID and a precise Description so the agent knows when to use them.

  3. Define Assistance Set the keywords or situations (such as "complaint" or "incorrect charge") that should trigger an alert for your human team.

  4. Test in the Simulator Before connecting WhatsApp, use the "Test Agent" tab. Pretend to be a difficult customer and tweak the instructions based on the results.