Automate intake and logging first; keep judgement and exceptions with a named human. That is the whole philosophy, but the reason it matters on WhatsApp specifically is tone. Customers treat WhatsApp as a personal channel. A bot that answers email-style questions in a rigid menu feels acceptable on a website; on WhatsApp it feels like being put on hold by a friend.

My earlier piece on automating a WhatsApp workflow without losing control covered the control design. This one is about the customer’s experience of that design.

What to automate without anyone minding

Acknowledgement that a message arrived and when a person will respond. Capturing structured details — order number, invoice reference, document type — so the human does not have to ask again. Routing to the right team. Sending status updates the customer asked for. Logging every conversation against the customer record. None of this pretends to be a person, and customers generally appreciate speed on routine items.

What must reach a human quickly

Complaints, anything involving money disputes, anything emotional, and anything the bot has misunderstood twice. Meta’s WhatsApp Business Messaging Policy says businesses may use automation when responding within the customer-service window but must also have prompt, clear and direct escalation paths — such as transfer to a human agent, a phone number or email. Design that handover as a first-class feature, not an error state.

The handover itself is where care is won or lost. The human should see the full conversation and the structured fields already captured, and should open with something that proves they read it. Nothing undoes goodwill faster than “Please share your order number” after the customer has already shared it with the bot.

A CA-practice example

Many CA firms in India already chase clients on WhatsApp for bank statements, invoices and sign-offs. A sensible automation acknowledges uploads, tags each document to the client and period, reminds politely on a schedule the partner approves, and flags anything unclear to the article or manager handling the file. It never answers a tax question on its own. Client documents are confidential and often contain personal data, so they should land in the firm’s approved document system rather than live in chat history, with access restricted. The DPDP Act, 2023 expects the business to protect personal data processed on its behalf with reasonable security safeguards; your processor contracts should reflect that.

WhatsApp’s policy also requires opt-in permission before you message people. Keep a record of how each contact opted in; it is a compliance point and a courtesy.

Measure the care, not just the speed

Track time to first human response on escalated threads, how often customers ask for a person, and how often the bot’s captured details were wrong. Read a sample of conversations every week. If customers repeatedly type “agent” or “call me”, your menus are in the way.

Related reading: start with the AI for chartered accountants in India hub, then the Monday morning test for any AI workflow.

Monday-morning checklist

  • Automate acknowledgements, data capture, routing and logging first.
  • Define escalation triggers and a visible route to a human.
  • Pass full context to the human so the customer never repeats themselves.
  • Record opt-in and move documents into your approved system.
  • Review a sample of real conversations weekly, not just response-time charts.

Sources

Educational content only — not legal advice. Check current platform policies and your data-protection obligations before deploying messaging automation.