Skip to main content
The AI Playground is the conversation simulator built into every agent edit screen in AI Agent Studio. Use it to sanity-check changes before they reach a real customer — type a question, watch how the agent replies, and see whether your latest brand information, personality, instructions, and skills produce the behavior you want.

​
Prerequisites

  • You already created an AI agent in AI Studio.
  • You completed at least the basic identity setup for that agent.

​
Where to find the Playground

  • Open the agent you want to test in AI Agent Studio.
  • On desktop, the AI Playground panel renders next to the editor on the right side of the screen.
  • On mobile and narrow screens, switch from Editor to Playground using the tab at the top of the page.
The Playground header reads AI Playground with a small subtitle that includes a select customer profile link.

​
Step 1: Send a test message

  • Type a message in the Playground input as if you were a customer (for example, “Do you have this bag in black?” or “Recommend a shoe for running.”).
  • Confirm the agent’s reply matches the identity, instructions, and skills you configured.
  • Send several follow-up messages to walk through a realistic conversation, not just one turn.
Test the common golden paths first (product question, order tracking, FAQ) and then try edge cases (out-of-stock items, unsupported topics, intentionally vague questions). Issues usually show up at the boundaries, not in the happy path.

​
Step 2: Personalize with a customer profile

The Playground can simulate a real contact so you can test how the agent personalizes replies.
  • Click select customer profile in the Playground header.
  • Pick a saved contact from your workspace.
  • The agent will now treat that contact’s saved data (name, language, fields you collect) as if it were a live conversation.
  • To go back to a generic simulation, click Reset customer profile.

​
Step 3: Reset the conversation between tests

  • Click Reset conversation in the Playground header to clear the message history and start a fresh ticket.
  • A reset clears the Playground state on both the front-end and the backend, so each test starts from the same point.

​
What the Playground does and doesn’t reflect

  • The Playground uses the currently saved identity, instructions, and skills for the agent. If you change a setting without saving, the Playground will not pick it up until you save.
  • Skill behavior (Create order, Order tracking, Product recommendation, Knowledge Base lookups, and similar) is exercised end-to-end, so tool calls hit real systems where they’re configured.
  • The Playground is the right place to catch behavior regressions before going live; it is not a substitute for monitoring real conversations once you launch.