Prepare a chatbot for website enquiries
Begin with a small set of questions that visitors regularly ask. Use approved business information and make it clear what the assistant can help with.
Create a short answer source
Collect the current service description, opening hours, delivery coverage, support contact and any public policies relevant to the page. Remove superseded versions. Keep private customer information and internal notes out of general visitor answers. Assign someone to update the source when the business changes.
Define the conversation scope
Write instructions that explain the assistant’s role, the facts it may use and what it should say when an answer is unavailable. Do not ask it to invent prices, confirm unavailable stock or promise an appointment without a supported workflow. Include a clear route to your published support contact when a person needs to review the enquiry.
Make the opening useful
Use a short first message that identifies the AI assistant and offers a few relevant topics. Ask only for information needed for the visitor’s request. Read the system-prompt and first-message guides, then configure the owned agent and integration after sign-in.
Test before a wider rollout
Try complete questions, misspellings, short replies and questions outside the source. Check the experience on a phone as well as a desktop. Review authorised conversation evidence in your own account and update unclear source material. Do not present a click on the chat button as a resolved enquiry.