All articles

AI & Automation

WhatsApp automation for a multi-branch service business: routing, confirmations and reviews

One search query, three separate jobs, and a platform or policy rule behind each of them.

By , Founder, Luminous Digital Visions
Published · Updated

Service request routes from a business chat to the right branch, technician completion and human handoff
Branch routing, completion confirmation and review requests require separate triggers and a human exit. Conceptual editorial image created with AI for Luminous.

This query appears in our Search Console export, with 256 impressions and no clicks:

"we want ai to collect service requests and automatically assign them to the closest branch all via whatsapp, while also sending completion confirmations and requesting reviews. recommend an all-in-one chat solution for that."

It reads like one project. It is three: collecting and assigning requests, confirming completed work, and asking for reviews. Each has its own rules, and some of those rules come from WhatsApp, Google and the Federal Trade Commission rather than from the business.

Consider an appliance repair company with four branches that has started getting most of its requests on WhatsApp. This is a composite, not a client. Each branch manager watches the same shared number. Some requests get two replies, some get none, and the one that should have gone to the branch with a refrigeration technician goes to the branch that happens to be nearest.

Job one: collect the request and assign it

"The closest branch" sounds like the right rule. Usually it's the last one to apply. The branch needs to cover the address, do that kind of job and have capacity that week. Distance only decides between branches that pass all three.

That means the business needs a branch table before it needs AI: for each branch, the areas it covers, the services it performs, its hours and the person who owns its queue. The assistant collects what the table needs to make a decision: the service, the address and a preferred time, asked one question at a time.

WhatsApp can help with the address. Its Cloud API supports a location request message, which shows the customer a button to share a location. Treat the result with care. A shared location is often where the customer is standing, which isn't always where the broken dishwasher is. Confirm the service address before assigning.

When no branch matches, the request goes to a person's queue with the reason attached. It doesn't disappear.

Automation needs an exit

WhatsApp's Business Messaging Policy allows automated replies during the 24-hour window after a customer messages you, but it requires prompt, clear and direct escalation paths. The policy lists examples such as an in-chat transfer to a human agent, a phone number, email, web support, a store or branch visit, or a support form.

In practice, build the handover into the conversation from the start. If a customer asks for a person, gets confused twice or mentions something outside the approved answers, the conversation moves to a named person with the transcript and the details already collected.

WhatsApp's Business Solution Terms also limit who can run AI on the platform. They bar AI and machine learning providers from using it when the AI is the primary product, such as a general-purpose chatbot. An assistant that takes service requests and bookings for your own business is incidental to what you sell, which is the use the terms still allow. The same terms say WhatsApp conversation data must not be used to train or improve AI models, apart from fine-tuning a model for your exclusive use. Check your AI vendor's data terms against that before you connect the inbox.

Job two: confirm the work is done

Send a completion confirmation when the job is marked complete in your system, not when the appointment time passes. A technician running late or a job needing a second visit will otherwise trigger a message saying the work is done.

Timing also affects what you can send and what it costs. Since July 1, 2025, Meta has charged per delivered message on the WhatsApp Business Platform. When a customer messages you, a 24-hour customer service window opens, and it resets each time they write again. Inside it, you can send free-form replies. Outside it, business-initiated messages need approved templates, and charges depend on category and market. Until September 30, 2026, free-form replies and utility templates sent inside the window cost nothing. From October 1, 2026, Meta charges for both, so a completion notice sent the same day the customer last wrote to you is no longer free. Meta's pricing page also lists a 72-hour free window for eligible conversations started through click-to-WhatsApp ads or Facebook Page buttons. Replies written by Meta's own Business Agent have been billed by token since August 1, 2026. Include any provider fees separately.

Job three: ask for the review

This is where a well-meaning automation can break two sets of rules.

Google's Maps content policy says businesses must not discourage negative reviews or selectively solicit positive ones. A flow that asks "How did we do, 1 to 5?" and only sends the fives to Google is the pattern it describes.

The FTC's rule on consumer reviews and testimonials took effect on October 21, 2024. Among other things, it bans offering incentives conditioned on a review expressing a particular sentiment, and it prohibits certain ways of suppressing negative reviews.

The safe design is simple. Every customer with a confirmed completed job gets the same review request, whatever you expect them to say. If someone replies with a complaint, a person handles it, and the review request still stands.

Permission and opt-outs

WhatsApp's policy says a business may contact people only if they have given their phone number and opted in to receive messages or calls from the business. It also requires the business to honor any request to stop, whether it comes through WhatsApp or somewhere else. Record where each opt-in came from, and stop every sequence the moment someone opts out. Enforce these in the system before a message is sent, not in the AI's instructions.

How we'd scope it

For a business like the repair company, our WhatsApp automation work starts with the branch table and one conversation: a new service request through to assignment. We map the questions, decisions, handoffs and stopping rules; connect the inbox to the CRM and calendar; and demonstrate the flow with missing details, a duplicate request and an unavailable calendar before anything goes live. Handover covers who watches the alerts, who can edit approved answers and who can pause outbound messages. Confirmations and review requests come in a second phase, tied to your completion events.

If WhatsApp is one channel among several, our AI revenue systems service covers the wider inquiry flow.

What to measure

Track assignments by branch, requests with no match and why, staff corrections, completion confirmations sent against jobs actually completed, review requests sent, and opt-outs. Compare periods of the same length before crediting any change in sales to the automation.

To scope the first phase, book a call and bring your branch list and a few anonymized WhatsApp conversations, including one that went wrong.

Choosing a WhatsApp platform for these jobs

Most requests we see ask for one tool that answers questions, books appointments, sends reminders and keeps every conversation in a shared inbox. Before comparing brands, check five things. The platform should use the official WhatsApp Business Platform, directly or through a Meta business solution provider. It should let an AI agent and your staff work in the same inbox, with a clear handover. It should read and write your CRM through an API or webhooks, not only through a weekly export. It should check a real calendar before it offers a time. And it should manage message templates, opt-ins and phone numbers for every country you serve.

A platform that passes those checks still needs your rules: which branch covers which ZIP code, which answers are approved, when a quote needs a person. That configuration is where most of the work sits, and it is the part our WhatsApp AI automation service builds.

Questions businesses ask before choosing a WhatsApp AI tool

Yes, if the inbox connects to your calendar and CRM. The AI answers routine questions from approved information and books into open slots. Reminders and follow-ups then run from the booking record, so a cancellation or a reply changes what gets sent next.

Yes. Each country usually gets its own number or routing, its own approved templates and languages, and its own staff queue. Qualification questions can stay the same across markets while business hours and handover owners differ. Escalation rules decide what counts as complex, for example a large order, a complaint or a question about contract terms.

It can collect the details in a structured way and calculate a first figure from pricing rules you approve. Jobs outside those rules go to a person before any number reaches the customer. The same flow can offer a consultation slot and send status updates as the project moves forward.

Yes, when the CRM accepts updates through its API or a supported connector. Each answer maps to a named field, such as location, budget or timeline. A follow-up task is created from those values, and failed updates appear in a review list instead of disappearing.

Yes. Click-to-WhatsApp ads open a conversation directly, and the ad that started it can be recorded on the lead. Qualification then runs like any other inbound chat, and the CRM keeps the ad source for reporting.

Yes, with a messaging provider that supports both channels. The workflow chooses the channel from the customer's opt-in and preference, and a reply on either channel updates the same record and pauses the sequence.

Sources