1. Create the workspace and agent
Start by creating a workspace for the business or client site you are implementing. Add one AI support agent for the brand, product, or support queue you want to automate. Configure its name, greeting, tone, fallback behavior, and human handoff expectations before adding it to a live site.
2. Add approved knowledge
Use the knowledge base to add help docs, policy pages, product information, FAQs, and URLs. The agent should answer from approved material instead of guessing, so keep source content current and remove outdated files when policies change.
3. Test before publishing
Use the playground to ask real customer questions, verify that answers cite the right knowledge, and adjust instructions where needed. Test edge cases such as refunds, pricing, account access, and questions that should be escalated to a human.