What Playground does
Playground is a safe testing area inside the Antbuildz AI Console. It lets you chat with the AI Agent as if you were a customer, without affecting real customer conversations.
Use it to test new setup instructions, uploaded knowledge, product information, scenarios, handover rules, and lead capture behaviour before launching changes to your live webstore, website widget, or WhatsApp channel.
Who should use this feature
Use Playground if you are setting up the AI Agent, updating business rules, adding new products, changing enquiry flows, or checking whether the AI is ready for customers.
When to use it
Use Playground before launch, after every important update, and whenever your team finds a weak answer in Conversations.
Before you start
Prepare 5 to 10 test questions based on real customer enquiries. Do not only test easy questions. Include normal enquiries, difficult enquiries, unclear questions, and situations where the AI should hand over to a human.
Example test questions:
- "Do you have a forklift for rent next week?"
- "What machine do I need for warehouse unloading?"
- "Can you quote me for a generator?"
- "Do you have spare parts for this model?"
- "I need urgent repair support today."
- "Can you confirm final price and availability now?"
Step-by-step guide
- Open Playground from the console menu.
- Select the AI Agent you want to test if your account has more than one agent.
- Start with a normal customer question.
- Read the AI response carefully. Check whether it is clear, polite, accurate, and useful.
- Continue the conversation like a real customer would. Give partial details, ask follow-up questions, or change your request.
- Check whether the AI collects the right information before creating a lead or handover.
- Test a situation where the AI should not answer directly, such as final pricing, unavailable stock, sensitive complaints, warranty claims, or technical advice that needs human review.
- Record weak responses. Decide whether the fix belongs in Agent Setup, Knowledge Base, Inventory, or Scenarios.
- Make the fix in the correct module.
- Return to Playground and test the same question again.
What good looks like
A good Playground result means the AI can handle the enquiry without confusing the customer or creating extra work for your team.
Good signs include:
- the AI understands the customer's intent
- the AI asks relevant follow-up questions
- the AI uses your business knowledge instead of guessing
- the AI collects contact details when a quote or follow-up is needed
- the AI hands over when the request needs human approval
- the AI does not promise final pricing, delivery, stock, or service commitment unless allowed
Example workflow
Your sales manager wants to confirm whether the AI can handle boom lift rental enquiries. In Playground, they ask: "I need a boom lift for building maintenance next Monday." The AI should ask for working height, location, rental duration, indoor or outdoor use, site access, and contact details. If it does that correctly, the workflow is closer to being customer-ready.
Common problems and what to check
The AI gives a generic answer. Improve Agent Setup with clearer role instructions and add better Knowledge Base content.
The AI cannot answer product questions. Check Inventory Management and Knowledge Base. The product information may be missing or unclear.
The AI asks too many questions. Simplify the scenario flow and decide which details are truly required before handover.
The AI does not hand over when needed. Add clearer escalation rules in Agent Setup or Scenarios.
The AI gives a confident answer when it should not. Add a rule that the AI must ask a human for final approval on pricing, stock availability, technical suitability, or urgent service commitments.
Done checklist
- You tested normal customer questions.
- You tested difficult or unclear questions.
- You tested at least one handover situation.
- You corrected weak answers in the right module.
- You retested after making changes.
Next recommended guide
Continue with Knowledge Base if the AI needs better facts, or Conversations if you want to review real customer chats after launch.
