WhatsApp chatbot examples: five useful patterns

Customer: "I need to change my delivery address." Bot: "Has this order shipped? I can help route the request."

A fixed menu works when each choice leads to a known next step. It does not need AI. Read an authoritative shipping state before allowing an address change; the customer's answer alone is not permission to edit the order.

If the answer does not match a choice, retain the original message and offer support. Measure whether the request reached the right team, not how many menu replies were sent.

A native WhatsApp Flow can collect structured answers such as a preferred date, service and location. A Logic Flow (Chatbot) coordinates the surrounding conversation and actions. These are different parts of the journey.

Customer: "Friday afternoon at the city branch." Reply after intake: "I have your preference. We still need to check availability."

The completed form is evidence of a request. A verified booking result is evidence of a reservation. If availability changes before confirmation, show a fresh choice instead of silently choosing another slot.

Customer: "Can I return this item after opening it?" The agent should use the policy that applies to that item, rather than a general return answer from memory.

The source needs an owner and an effective date. If the policy is missing or conflicting, the answer should explain that a teammate needs to check. Keep customer-provided claims separate from the store's rules.

Start with the knowledge-base preparation guide. The success condition is a supported answer to the actual question, including exceptions that matter.

Customer: "Where is order DEMO-204?" The workflow first establishes the authorised customer-order match, then reads the fulfilment record.

A result might say that one parcel shipped and another is still being prepared. Do not collapse that into "Your order has shipped." If the connection times out, say the status could not be checked. "Order not found" would be a different result.

This example is a design pattern, not evidence that every store connection is configured. The Shopify tracking guide explains the setup boundary.

Customer: "That answer is wrong. I need someone to check." Offer a direct staff route. Preserve the request, checks attempted and unresolved question.

The bot should yield while the person handles the conversation. During closed hours, state the real staffed hours and available alternatives. Do not say someone is reviewing the case merely because it entered a queue.

Read the handoff decision guide for escalation triggers. The useful outcome is an accepted next action, not an automatic acknowledgement.

Write four lines for your first use case: customer question; trusted source; permitted action; recovery when that action fails. A menu may be enough for routing. A form helps structured intake. AI helps interpret varied questions. Connected actions need explicit permissions.

Try a complete request, a correction and a missing-data case with fictional records. If you cannot explain the recovery, simplify the workflow before adding more automation.