ChatLik AI Playground — test before customers

AI Playground runs the same pipeline as real Messenger — full instruction, RAG, products, orders, and human handoff — without customer risk. Select the Page, start a New conversation, and use Quick Prompts: Greeting, Products, Order, Human.

Before you start

1. Save Business Profile, FAQ, products, and settings for the Page under test.
2. Check credits — Playground consumes them like live chat.
3. Open Playground from ChatLik AI Preferences or its tile.

Run a test session

1. Select the Page inside Playground.
2. From the Chats sidebar start a New conversation.
3. Type in Type a message... or use Quick Prompts: Greeting / Products / Order / Human.
4. Review the reply and stats (Model / Tokens / Time) when shown.
5. Test an image if Receive Images in Chat is enabled.

Manage chats and verify

1. Copy the full chat to share with the team when needed.
2. Clear messages or delete the conversation from the sidebar after testing.
3. Mandatory go-live scenarios: price, out of stock, returns, out of scope, human request.
4. If human handoff fires, confirm the row in Live Support.

After setup and verification

1. Document for the team: selected Page, who owns review, and when Playground was last tested.
2. Review day-one live replies, any error logs, and unexpected credit usage.
3. If price, policy, or hours change the same week, re-run the critical scenarios before trusting automation.
4. Keep a clear human escalation path when automation fails — never leave the customer without a fallback.

Ongoing quality control

1. Assign one owner to review this screen weekly so responsibility is not diffused.
2. On repeating customer complaints, reopen related settings and compare them with Business Profile and FAQ.
3. Log pause/edit decisions with a date so the next teammate knows what changed and why.
4. Do not treat setup as finished after the first save — review it with every new offer or season.

Practical tips

• Retest on the production Page after copying settings from a sandbox.
• Log failures with the reply text for the team.
• Retest after every major catalog or instruction change.
• Recheck permissions if Save/Refresh buttons disappear after a role change.

Common mistakes

• Going live without a Human prompt test.
• Trusting a good reply while the wrong Page is selected.
• Deleting the failing chat before sharing it with whoever fixes prompts.
• Treating Save alone as success without a realistic customer scenario test.

Related: Agents · Products · Module map.

Next step

Try ChatLik via sign up or compare plans.