WhatsApp chatbot for an SME: the steps in order (without skipping the basics)

For many SMEs, WhatsApp is where sales and support really happen. That makes it a high-impact channel for an AI assistant — and a high-risk one if you skip infrastructure or launch without clear rules. Here is the sequence we follow, in order.

Step 1: Confirm you need the Business API

An AI agent does not connect to the WhatsApp app on a company phone. It needs the WhatsApp Business API, managed through an authorized provider. Automating on top of the regular app risks number blocks and offers no reliable way to scale. This is not a detail you can fix later; it is the foundation.

Step 2: Map real conversations, not ideal ones

Before writing a single reply, collect 20 to 30 actual customer messages from the last month: WhatsApp, email, DMs. Sort them into what is repeatable (hours, prices, availability, order status) and what needs a person (complaints, negotiations, anything emotionally charged). The agent handles the first group; the second group defines your escalation rules.

Step 3: Define tone, hours and the first message

The agent should sound like your team, not a generic chatbot template. Outside business hours, it can still answer simple queries, but it must be explicit about when a human will follow up — otherwise customers assume someone is “there” at midnight.

Watch out: launching without hour rules is the fastest way to get complaints that the bot “does not help.”

Step 4: Build the knowledge base from verified answers

Every answer the agent gives should come from content your team has already validated: FAQ doc, price list, policy pages. Guessing or letting the model improvise on pricing or legal terms creates liability and erodes trust the first time someone checks.

Step 5: Connect the tools that matter

At minimum, decide where escalated conversations land (email alert, CRM ticket, shared inbox). If the agent qualifies leads or books appointments, connect calendar and CRM early so nobody has to re-type what the customer already said.

Step 6: Test with internal users, then a soft launch

Run real scenarios with your team before customers see it: wrong questions, angry tone, “I want to speak to someone,” edge cases in your product catalog. Then enable it for a subset of traffic or one entry point (website widget linking to WhatsApp, QR on invoices) before going site-wide.

Step 7: Monitor and adjust for the first weeks

Track conversations resolved without human help, time to escalation, and repeat questions (a sign the first answer was not enough). The first month is tuning, not “set and forget.”

How we do it at Bitora

We set up the WhatsApp Business API correctly from day one, design tone from your team’s real messages, and wire escalation and CRM before the first customer-facing reply. Data is processed on our own infrastructure in Spain (EU), with encryption and processor agreements where GDPR applies.

If WhatsApp is a core channel for you, see the step-by-step offer on chatbots for business or business chatbot solution.