Skip to main content
Waaru
WhatsApp human handoff

When should a WhatsApp chatbot hand over to a human?

Tanay PatelAI and full-stack engineer
The short answer

A WhatsApp chatbot should hand over when the customer asks for a person, the answer lacks a reliable source, or the request needs a decision the bot is not authorised to make. Failed actions and repeated misunderstandings also need an exit. A useful handoff tells the customer what happens next and gives the receiving teammate enough context to continue.

Decide these conditions before writing a welcome message. A bot that collects the right details can still leave a customer waiting if nobody owns the next step. This guide helps you choose escalation rules for a rule-based chatbot or an AI agent. The examples are illustrative, not reports of customer results.

Which requests should go straight to a person?

Start with the customer's explicit choice. If someone asks to speak to a person, provide that path without making them finish a sales questionnaire. Then list the decisions your team reserves for people. These might include a custom commercial quote, a disputed charge, or an exception to a published cancellation policy.

A routine policy question can stay with automation when there is a current, applicable answer. Deciding whether to waive that policy is a different job. Name the team authorised to make that decision.

  • Requested person: acknowledge the request and route it with the conversation so far.
  • Missing or conflicting facts: state what cannot be confirmed and ask the responsible team to check.
  • Human approval required: collect only the details needed for the decision. Do not promise approval.
  • Failed action: record what was attempted and its known result. Do not tell the customer it succeeded.
  • Repeated misunderstanding: offer a person instead of cycling through the same clarification.

How do you turn those situations into handoff rules?

Write each rule so a teammate can test it with a sample message. 'Hand over when appropriate' is too vague to check. 'When the customer requests a refund outside the published policy, collect the order reference and send the request to support' identifies a condition, an input, and an owner.

Use this worksheet for each situation. Decide how your chosen software will detect the condition; an AI model's statement that it is confident is not evidence that an answer is correct.

  • Condition: what customer request or failed step causes the handoff?
  • Evidence: which message, missing source, or action result shows the condition occurred?
  • Response: what can you truthfully tell the customer at this point?
  • Destination: who checks the queue, and who covers their absence?
  • Stop rule: which customer-facing automation must pause while a person handles this request?
  • Completion: how will the team know someone has accepted and resolved it?

What should the customer hear during a handoff?

Keep the acknowledgement specific. Explain the unresolved task and where it is going. Only say a teammate has joined when that has actually happened. Sending a request to a queue is an earlier step.

Illustrative order enquiry: Customer: 'The tracking link does not work. Has my order shipped?' Bot: 'I could not confirm the shipment status. I have sent your order reference and this conversation to support for checking.'

Illustrative sales enquiry: Customer: 'Can you give us a custom rate for several locations?' Bot: 'That needs a sales review. Which locations should the team include in the quote?' After routing succeeds: 'I have sent the request to sales. A teammate will reply here.'

If routing fails, the bot must give the actual available support path instead of claiming that a team received the request. Avoid a promised reply time unless your staffed process can support it.

What if your team is offline?

Give the team's actual staffed hours and explain whether the enquiry has been saved for review. An acknowledgement should not imply that a live agent is waiting. Decide who checks overnight enquiries at the next staffed period, including requests that arrived before a holiday.

Example wording to adapt: 'Your enquiry is in the support queue. The team is currently offline and will review it during the staffed hours shown here.' Replace 'shown here' with your actual hours and timezone before using it. Offer an alternative contact method only if someone monitors it.

Keep the original conversation available. Asking the customer to start again by email may be necessary in some cases, but it should be a deliberate support choice with a clear reason.

What context does the teammate need?

Pass the customer's request, facts already supplied, the unresolved question, and any action attempted. Distinguish an observed result from an assumption. 'Order lookup timed out' is useful; 'order is probably delayed' is an unsupported conclusion.

The conversation history is the evidence. A summary, if your system supplies one, is a convenience that a person should be able to check against that history. Do not assume every handoff includes an AI-generated summary. Keep internal notes out of the customer-facing reply and share only information the receiving team needs.

Which WhatsApp rules still apply after handoff?

WhatsApp permits automated responses during the customer-service window while requiring a clear escalation path. An in-chat transfer is one option; its policy also lists alternatives such as email and a support form. A human reply does not bypass the messaging window. Outside 24 hours of the last customer message, the Business Platform requires an approved message template. Check the current WhatsApp Business Messaging Policy when designing your queue and follow-up process. Consent and opt-out requirements still apply.

How can you test a handoff before customers use it?

Run sample conversations through the full path. Check the destination inbox as well as the customer's view. A message saying 'transferred' does not prove anyone received it. Record the expected outcome and the actual result for each case, then correct failed cases before publishing the workflow.

  • Ask for a person at the start and midway through the conversation.
  • Ask a question the approved information cannot answer. Check that the bot does not invent an answer.
  • Simulate an unavailable lookup or action. Check that an uncertain result is not presented as success.
  • Repeat a misunderstood request. Check that the customer can leave the clarification loop.
  • Make the intended teammate unavailable. Check who receives the queued enquiry and what the customer is told.
  • Accept the conversation as a person. Check that the bot does not continue replying to the same request.

Where should you start?

Choose one recurring exception and give it an owner. Use the WhatsApp human handoff guide to understand conversation context and assignment, then check your inbox setup. If missing business information causes the handoffs, work through the chatbot knowledge-base checklist before expanding what the bot answers.

FAQ

Frequently asked questions.

Does human handoff require an AI chatbot?

No. A fixed workflow can hand over at a configured step or fallback. AI can help interpret varied requests, but the receiving team, context and ownership still need to be defined.

Should every unanswered question go to a person?

A short clarification can help when a required detail is missing. If there is no reliable answer, a decision needs human authority, or clarification keeps failing, provide a person or another monitored support path.

Does a human handoff reopen the WhatsApp 24-hour window?

No. Transferring ownership does not reset the customer-service window. The last customer message determines it, and replies outside the window require an approved template under WhatsApp Business Platform rules.

What should a handoff summary include?

Include the request, confirmed facts, steps attempted and the unresolved decision. Keep the source conversation available. If a summary is generated by AI, check that it has not converted an assumption into a fact.

Research notes

Evidence checked for this article.

  • WhatsApp requires clear escalation paths for automated responses and sets the customer-service window and template rules.

    WhatsApp Business Messaging Policy · Accessed 6 September 2026

About the author

Tanay Patel

AI and full-stack engineer

Tanay builds production AI systems, agent workflows, and the full-stack software around them. His work focuses on reliable state, recovery, and human handoff.

Read Tanay's engineering notes

Put this into practice on your own number.

The platform is free. AI features are paid separately.