WhatsApp AI agents for sales: from enquiry to human handoff

Start with one enquiry type your team already handles. Write down the decision the conversation should support. For an agency, that might be whether the requested work fits its services and who can decide what work is needed.

Keep fixed rules where they are useful. A service menu or required location check does not need AI. AI can help interpret a loosely worded request or draft an answer from reviewed service information. Use the chatbot versus AI agent guide to choose the approach for each step.

Customer: “We need help with WhatsApp enquiries. Can you send a quote?”

Suggested reply: “Which part would you like help with: answering enquiries, booking requests or follow-ups?”

Customer: “Booking requests. People ask about dates and then stop replying.”

The next question should clarify the work: “How does your team check availability today?” If the customer already described their setup, use that information rather than asking again. The lead qualification guide helps you choose the wording.

Suppose the answer reveals a booking system the agency has not evaluated. The useful response is: “The team needs to check that booking system before pricing the work. Would you like to discuss it with them?” A plausible-sounding integration promise would create more work for everyone.

Use this worksheet to decide who handles each step. Note where the information comes from and what the customer should hear if something fails.

Download the worksheet (CSV)

Open it in Excel or Google Sheets. Each step has a labelled example and blank fields for your own workflow. The examples describe a fictional agency enquiry.

Use a person when the scope, price or delivery date needs judgment. In this example, the booking system has not been assessed, so the workflow prepares the enquiry for review. It does not invent an implementation cost.

A published price list can support an answer about a standard service. Still check what that price includes and whether the customer’s request fits. If a tool returns a price, verify its source and conditions before presenting it as a commitment. The tools and actions guide explains the difference between reading information and changing a business record.

Tell the customer what is happening. “I have passed your booking-system question to the team” is appropriate only after the transfer succeeds. If it fails, explain that the request has not reached the team and offer another working way to contact the business.

Give the teammate the original context as well as any summary. A summary can omit the detail that changes the quote. Stop automated sales replies while that person owns the conversation, according to the workflow’s handoff rules.

If the team is offline, state its working hours or when the customer can expect a reply. Do not promise an immediate reply. Use the handoff decision guide for reply examples and what to do when a transfer fails.

Agree on the next step while the customer is present. An enquiry is not blanket permission for unrelated marketing. The WhatsApp Business Messaging Policy requires opt-in for subsequent messages and respect for opt-outs.

Outside the 24-hour customer-service window, WhatsApp requires an approved template. Passing the conversation to a person does not remove that rule. Use the opt-in guide to plan permission and keep the follow-up relevant to the agreed purpose.

Run the conversation on paper first. Try a clear request, missing information, a changed requirement and a customer asking for a person. Check whether the receiving teammate can take over without starting again.

Then use the AI-agent test guide to record what you expected and what happened in your chosen system. A good result is a useful enquiry that reaches its owner. Sending more messages is not the goal.

The agency guide explains how to assign enquiries to the right person. For the fixed steps, start with creating a Logic Flow (Chatbot) and confirm the capabilities available in your workspace before adding AI or external actions.