Test in the playground

The Playground sits on the right side of every AI Employee card (or opens full-screen on smaller windows). It is a live chat with the employee, using whatever settings are currently saved across every tab, and nothing sent here reaches a real channel.

​

Why the Playground matters

Every setting across all seven tabs of the AI Employee configuration ultimately shapes how it behaves in an actual conversation — the Playground is where all of that comes together and gets verified directly, rather than trusted blindly. Use it to check:

• Does the tone of voice actually come through the way you intended?
• Are boundaries actually being respected — does the employee decline the topics you marked off-limits?
• Is Conversation flow & objectives correctly triggering the right content, actions, or assets at the right moments?
• Does Identity & style produce the personality you were aiming for?
• Do exit points fire when they should — does asking for a human actually trigger the handover?
• Are assets actually sent when the relevant condition comes up?
• Is content actually informing accurate answers?

​

Pick a test contact first

Select a test contact from the dropdown before sending a message. Without one, the employee has nothing to read, share, or update, so any guideline or action that depends on contact data will not behave the way it does live. Once selected, the conversation pulls that contact’s real data — first name, email, phone, custom attributes — and updates it live as the conversation progresses, if the employee has Update permission turned on for that field.

​

The session persists across tabs

Send a message, then switch to another tab — Content, Attributes, Actions, whatever you need to adjust — and the conversation stays exactly where you left it. Tweak a guideline, add a knowledge source, or toggle an attribute, then jump back to the same chat to see the effect immediately, with no need to start over.

​

What you can send

The Playground supports text, images, video, documents, and emojis. Voice is not supported here yet, though it works on live channels — that is a Playground-specific gap for now.

​

Starting clean

Click Reset chat to clear the conversation and start from scratch, with no history carried over. Use this whenever you are testing one specific scenario and do not want earlier messages influencing the reply.

​

Before you go live

Test the situations you are most worried about, not just a friendly hello:

• Ask something that should trigger an action, and confirm it fires correctly.
• Ask something outside a boundary, and confirm the employee holds the line — boundaries are instructions, not a hard filter.
• Push toward an exit point — frustration, a refund request, whatever you have defined — and confirm it hands off the way you expect.

Once you are happy with how the employee behaves, click Set Live to activate it — only live employees can be added to an automation. If you need to pause one without deleting it, click Set Inactive instead; it stays in your list and keeps its setup, it just stops running until you switch it back on.

​

Best practices

• Test after every meaningful change, not just at the end of a long editing session — it is much easier to identify which specific change caused an unexpected behavior if you test incrementally.
• Test edge cases deliberately, not only the obvious happy-path questions. Phrase a question awkwardly, bring up a boundary topic directly, ask for something the employee should decline, and confirm the response matches what you configured.
• Test from the customer’s actual perspective, using the language and tone real customers tend to use, rather than clean, formal test phrasing that does not reflect how conversations unfold in the wild.
• Revisit the Playground after launch, not just during initial setup — especially after adding new content, assets, or actions, to confirm the additions integrate correctly with everything already configured.