How to integrate WhatsApp with Odoo without losing traceability

The scene repeats itself at almost every midsize company in the region: 60% of opportunities come in through WhatsApp, none of the data from those conversations reaches the CRM, and when a salesperson quits, the company loses the entire history of their customers.

What exactly is being lost

• Source. Nobody knows which campaign or referral produced the inquiry, so you can't measure what works.
• History. What was promised to the customer lives in a private chat.
• Follow-up. Unanswered opportunities don't show up in any report because they don't exist in the system.
• Response time. It isn't measured, and on this channel it's the factor that most affects closing.
• Continuity. If that person isn't around, the customer starts from scratch.

The architecture that works

1. A company number, not a person's. It's the decision that makes everything else possible.
2. Every incoming conversation creates or updates a record in Odoo: an opportunity if it's a prospect, a ticket if it's a service request, a contact if one doesn't exist yet.
3. The message is saved to the record's history, so anyone authorized can see the full conversation.
4. Salespeople reply from wherever is convenient — inside the ERP or from the app — and the reply is logged either way.
5. The rules live in the ERP: assignment by shift or territory, inactivity alerts, escalation when a message goes unanswered too long.

Technically, the connection is made against Odoo's API — JSON-RPC — from the service that receives the messages. It's the same pattern we use to run our own messaging infrastructure in production.

Five common mistakes

1. Automating replies before fixing the process. A bot that responds quickly to a messy process only speeds up the frustration.
2. Saving only the text. Without the link to the customer, the order and the person responsible, the record is useless for decision-making.
3. Not setting business hours. If the channel is open 24/7 and the team answers 8 to 5, that has to be communicated automatically.
4. Letting the automated agent make promises. Prices, lead times and availability must come from ERP data or be escalated to a person.
5. Forgetting about consent. Sending bulk messages without prior opt-in damages the brand and exposes the company.

What you can automate next

Once the channel is connected, the next steps follow naturally: automatic order confirmation, shipping notices, payment reminders with the account statement, and availability questions answered with real stock levels. That's the point where an AI agent starts to make sense, because it has reliable data to query.

The general method is in systems integration.