The problem.
Javi runs operations at a tech services company with over 200 active customers. His support team — four technicians — received tickets through two channels: a shared email mailbox and a WhatsApp Business number. Average volume: 15-25 incidents/day, with peaks of 40 on bad days.
The problem wasn't volume. It was the absence of criteria. When a customer wrote by email, any of the four technicians could see it — but there was no way to know who had read it, who was working it, or whether it was already resolved. Tickets duplicated (customer writes by email, doesn't get a quick reply, writes by WhatsApp). Two technicians could be working on the same thing without knowing. And a critical server outage could end up buried beneath a question about how to change a password.
Javi had already tried a traditional helpdesk. The team ignored it: they kept replying directly via email and WhatsApp because it was faster, and the ticket system fell out of sync. He needed a solution that adapted to the team, not the other way around.
What we built.
Invisible classification layer
An n8n workflow listens to the email mailbox and the WhatsApp Business number. Each incoming message goes through a Claude classifier that extracts: customer, category, priority, affected product, and detected sentiment. Records to PostgreSQL with a unique ticket ID, but without forcing the team to use anything new.
Cross-channel deduplication
If the same customer writes via email and then WhatsApp about the same thing, the system detects it and unifies into a single ticket. Zero duplicates processed twice. The technician sees the merged history.
AI auto-resolution for L1
For repetitive tickets (password reset, basic configuration, where to find X), the AI replies directly with the documented solution from the knowledge base. The technician validates with one click. 40% of L1 tickets close without human writing.
Slack-integrated dashboard
The team sees the prioritized queue in a dedicated Slack channel: pending tickets sorted by SLA risk, who's handling each, age, status. The team didn't change tools — they got a clean view of what already existed.
The team's reaction was "wait, this just works?" Yes. They keep replying by email and WhatsApp like always. The system organizes everything around them.
Stack used.
n8nClaude SonnetWhatsApp Business APIGmail APIPostgreSQLSlack