Inbox that manages itself

The context

In B2B consulting and outsourcing companies, the operations team manages communication with clients, vendors, and internal teams through a shared inbox that receives between 180 and 250 emails per day.

The typical process is entirely manual: every morning, two team members spend the first 3 hours reading each email, deciding if it's urgent or routine, forwarding it to the right department, and replying to those with standard answers. They use Gmail with manual labels and a spreadsheet for tracking.

The challenge

Email volume grows every quarter, but the team doesn't. What starts as a 1-hour task becomes half a day's work. Urgent emails — client complaints, service incidents, deadline-sensitive requests — get buried under newsletters, auto-confirmations, and routine inquiries.

The cost isn't just time. Delays in responding to urgent emails cause unnecessary escalations, unhappy clients, and in some cases, contractual penalties. The team ends up spending their best energy on a mechanical task instead of solving real problems.

The solution

The system is built with n8n and the OpenAI API, connected to the Gmail inbox via API. Each incoming email goes through a classification pipeline: the AI model reads the subject and body, identifies the email type (urgent, routine, informational, spam), and labels it automatically.

Routine emails with standard responses — receipt confirmations, basic information requests, status follow-ups — are replied to automatically with smart templates that adapt tone and data to the original email's context. The team reviews and approves all templates before launch.

Urgent emails are immediately forwarded to the right person with an AI-generated summary that includes: who's writing, what they need, urgency level, and suggested action. The team gets a Slack notification with the summary so they can act within minutes. Typical implementation takes 4 days.

Results

Daily time spent managing the inbox drops from 3 hours to 15 minutes. Those 15 minutes are used to review emails the system flags as "requires human decision" — less than 10% of the total.

Urgent emails get handled in an average of 8 minutes, down from 2-3 hours with the manual process. Delay-related escalations drop by 90% in the first month.

The operations team recovers half a day of productive work. It's like having an assistant that never gets tired, never sends to the wrong person, and never forgets to reply.

Lessons learned

• Not every email needs AI. 60% of emails are resolved with simple classification rules. AI is reserved for ambiguous cases.
• Involving the team in defining response templates is key. They know the right tone for each type of client.
• The automatic summary of urgent emails is the most valued feature. It doesn't just save reading time — it improves response quality because the team arrives at the email with context.