Build your first bot flow

The most common mistake is a decision tree that tries to handle everything. It takes a week, customers fall out of it in ways you did not anticipate, and you turn it off.

Start with one job the bot can finish. Add the second only after the first has run for a week.

Good first flows

  • Opening hours and location. Asked constantly, never changes, needs no person.
  • Order status. Reads the real order from your store, so the answer is true rather than a holding message.
  • Menu or catalogue. Send the list, let them pick, take the order.
  • Qualify an enquiry. Ask the three questions your team always asks, write the answers onto the lead, then hand to a human who arrives already knowing the budget and the dates.

Building it

  1. Open Bots → New flow and name it something you will recognise in six months.
  2. Set the trigger: a keyword, a button on another flow, the first message of a new conversation, or a QR code scan.
  3. Add the steps — a message, a question, buttons, a list, an image, or a condition on something you already know about the customer.
  4. Put a handover in. Every flow needs a way out to a human: a "talk to someone" button, and an automatic escalation when the customer says something the flow does not understand twice.
  5. Test it on your own number before publishing. The testing panel replays the flow exactly as a customer sees it.
  6. Publish.

Where people go wrong

  • No way out. A customer stuck in a loop with no human to reach will not message you again. This is the single most damaging bot mistake.
  • Too many questions before any value. Three is a conversation. Seven is a form, and they leave.
  • Buttons that overlap. "Order" and "Place an order" as two buttons means half pick the wrong one.
  • Writing like software. "Invalid input. Please select a valid option." is how you sound like a bank's phone menu. Write what a good member of staff would say.

Bot or AI?

A flow is right when the path is fixed and you want the same outcome every time — ordering, booking, qualifying. AI is right when the question is open and the answer lives in your catalogue, prices or FAQs.

Most businesses end up with both: AI answers the open questions and hands into a flow when the customer wants to actually do something. See AI agents.

Knowing whether it works

After a week, look at three numbers in analytics:

  • Completion rate — how many who started finished. Under half means a step is losing them.
  • Handover rate — how many asked for a person. High is not failure; it tells you what to build next.
  • Response time — on the conversations the bot handled, against the rest.

If completion is low, open five real transcripts and read them. The step people drop at is usually obvious within a minute, and no dashboard tells you as fast.