Every hour the system reads the new emails in the shared inbox, understands what they are about and assigns a category. The email lands in the right folder and in the right department's inbox, and a summary of every decision arrives on Slack. Nothing goes out to the customer automatically.
The info@ inbox is the one everybody knows, so it receives everything: quote requests, supplier invoices, job applications, support requests, sales pitches, spam. Someone opens it every morning, reads, decides whose it is and forwards it.
While the emails are few, it takes a minute. As they grow it becomes part of someone's job, one that stops when that person is on holiday or in a meeting. And an email forwarded to the wrong department sits there until someone notices.
An n8n workflow checks the inbox every hour and picks up only the new emails. There is nothing to install on anyone's computer, and the inbox stays the one you already have.
The model reads subject and body and picks a category from a closed list agreed together: sales, administration, support, applications, other. An invoice sent by a supplier's sales rep goes to administration, not to sales.
The email is moved to its category's folder and forwarded to the right department's inbox. Whoever handles it finds it where they already work, without changing tools.
For each email: sender, subject, category and department. If a decision is wrong it shows at once, gets corrected and becomes an example for refining the instructions.
These are the numbers we set in discovery, before writing any code. The starting value is measured on your data, and the prototype is judged against it.
The share of emails that end up in the right category, checked against the Slack summary. It is the number that decides whether the system can work on its own.
The hours someone spent every week opening, reading and forwarding, before and after.
How long passes between an email arriving and the right person picking it up. It is the number the sender feels.
If the inbox gets few emails a day, or almost all of them go to the same department, a mail client rule is enough and costs less. And the system replies to nobody: if the problem is how long customers wait for an answer, this is the first step, not the solution.
Delivery notes, supplier confirmations, contracts, certificates. AI recognises every document, extracts the data for the ERP and flags what a person needs to look at.
Customers and colleagues ask questions, and the assistant answers using only your documents and the pages of your website. Every answer comes with its source, and when it does not know, it says so.
A free 30-minute call, with no commercial follow-up unless you want it.