Train your assistant

How Wabo learns

There is no "training" in the technical sense. Each time a customer writes, Wabo sends Google Gemini the conversation together with your catalog, business details, rules and the relevant parts of your documents. Better input gives better answers, and every change you save is used from the next message. Start with your products (see importing from Excel), then work through the steps below.

Step by step

1. Describe your business. Open Train the agent. In the Business and tone card, fill in Business name, choose a Conversation style (Friendly, Professional or Energetic), and write About your business: what you sell, where you ship from, delivery, returns. Add a Greeting example in your own words, because the assistant copies its tone. List your Payment options, then click Save.
2. Add info, rules and FAQs. Use Add new info, rule or FAQ. Info is for facts, Rule for behaviour, FAQ for a Question and its Reply. Click Add. Entries are listed under What the agent knows, where you can Deactivate or Delete them.
3. Upload documents. In Documents, click Add document or Paste text.
4. Choose how it sells. In Settings, under What should the agent do?, set Product sales.
5. Test in Test chat. Ask the questions your customers really ask.

Good info and rule examples

Short, specific entries work best. One topic per entry makes them easy to update.

• Info: "Orders placed before 2 pm ship the same day. Delivery takes 1–3 business days."
• Info: "Returns are accepted within 14 days if the product is unused and in its box."
• Rule: "Do not negotiate on price. If someone asks for a discount, hand the conversation over to us."
• Rule: "Never promise a delivery date for custom orders; say we will confirm it."
• FAQ: Question "Do you have a physical store?" Reply "We sell online only, with shipping across the country."

Avoid vague rules such as "be helpful" or "sell more". They do not change anything. If a rule matters, describe the situation and the exact behaviour you want.

Documents

Price lists, service brochures, return policies, catalogs and FAQ sheets can go in as documents. Supported formats are PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), text (.txt, .md, .csv) and web pages (.html). The text is extracted on your computer; when a customer asks something, only the relevant sections are sent to the AI. Scanned (photo) PDFs cannot be read.

Use Try it: which section would be found if a customer asked this? to check: type a question like "How much is shipping?" and click Search. If the right section does not come up, the document probably needs clearer wording, or the fact belongs in an Info entry instead.

Handover or order mode

Under Product sales in Settings you choose one of three:

• Off: for businesses that do not sell products, for example appointment-only. The assistant answers questions and passes buyers to you.
• Handover: greets, sends the price list and, if asked, product images, then leaves the conversation to you. With After sending product images, let the agent go quiet and I'll take over the conversation ticked, it stops replying in that chat after the images.
• Order: collects product, option, quantity, name, phone, address and payment method, has the summary confirmed and saves the order. You can be notified of each new order on your own WhatsApp.

Testing in Test chat

Test chat works without WhatsApp and without a license. Write as a customer, or click one of the suggested questions. You can attach an Image or Audio file to try photo and voice messages. Nothing is sent to WhatsApp and no order is saved; instead you see notes such as Handed over to a human or An order was about to be saved, with the response time and token count under each reply.

If you have more than one AI connection, use Compare side by side… to send the same message to two models at once. It is a quick way to decide whether a cheaper model is good enough for your shop. Click New chat to start over. To add a second connection, see the Gemini API key guide.

When the answers look right, link WhatsApp and turn the agent on.