AI answers from your knowledge, with your team ready to take over.

Start with the information you want customers to receive

In your project’s AI settings, write a knowledge base with your product information, common questions and current policies. Specific answers work better than a list of topics: explain what customers can do, any conditions that apply, and when they should contact your team.

The assistant uses this knowledge when composing business answers. Its instructions ask it to request a person when the relevant information is missing, instead of assuming a price, policy or account detail. You can also set the handoff message customers receive.

If your information already lives in documents, activate the separate Knowledge files module. It accepts Markdown, plain text, CSV and PDFs with extractable text. Suppy extracts their text for the assistant to use alongside your written knowledge base. A scanned PDF without readable text needs to be exported to a supported text format first.

Keep questions and requests for action distinct

Explaining a policy and carrying out a customer request are different jobs. The assistant is instructed to answer a policy question when your knowledge covers it. Requests to change an order or account, such as processing a refund, are directed to an operator. Missing business information and an unusable AI response can also lead to human help.

Operators can take over from the dashboard. Takeover switches that conversation to human handling and turns off AI replies for it, while keeping the existing messages together. Your team can return the conversation to AI handling when appropriate. Handoff does not guarantee that an operator is immediately available; write your handoff message to match the support your team can provide.

Try a complete conversation in three steps

This example is illustrative: imagine a small online shop that wants to answer product-care questions and route individual order changes to its team. The sample policy below is not a Suppy policy or a claim about a real customer.

1. Prepare and enable. Activate the AI assistant module, using an available trial if offered, and enable the assistant in your project’s settings. Add a sample rule such as “Wash the cotton tote on a cold cycle and air-dry it.” Add an escalation message that explains a team member will need to handle order changes. Then install the widget on your test page.
2. Test a covered question. Ask “How should I wash the tote?” Check that the response reflects the supplied instructions without adding unsupported details. Try variations in wording, then correct unclear or conflicting knowledge before launch.
3. Test the handoff. Ask “Please change the address on my order.” Check that the chat requests an operator. Take over in the dashboard, read the preceding messages, and reply as a person. Also test a business question absent from your knowledge base.

Choose modules for the work you need

The AI assistant is an optional project module. Knowledge-file uploads and chat attachments have their own modules; they are different from writing directly in the knowledge-base field. See current seats, module prices and trial availability before activating them. Where a trial is available, it is offered once per project and module; continuing after it ends requires a separate purchase.

Keep the knowledge base current as your products and policies change. AI can make mistakes, so review important questions and keep an operator path available. For customers who prefer another language, explore live chat translation and test your key terms in the languages your customers use.

Questions about the assistant