Skip to content

Automation / Symfony + AI

A mail sorter for a hotel chain: 95% of emails without reception

A hotel chain received hundreds of emails a day across its reception inboxes, a shared inbox and booking portals. We built a sorter that recognises each message and sends it to the right hotel and department. 95% of the mail reaches its place with no reception involved.

Mail handling automation

95%
of mail sorted automatically
with no reception involved; only what needs a decision reaches a person

Starting point

Hundreds of emails a day, split up by hand

Every hotel in the chain has its own reception inbox, with a shared inbox above them all. On top of that come bookings from external portals. Hundreds of messages arrived every day: bookings, date changes, group requests, invoices and spam. All of it landed on reception desks with no order. Someone had to read each email, work out what it was about and pass it by hand to the right hotel or department. Before a message reached the person who should deal with it, it went through someone else's hands, and at that volume the work never ran out.

What we did

Rules from hotel staff, AI for the uncertain cases

The sorter assesses every email in two ways. Rules recognise most of the mail, and an AI layer handles only what the rules are not sure about.

  1. 1

    One entry point for every inbox

    The sorter takes in mail from the shared inbox, from each hotel's reception inbox and bookings from external portals. It runs on Symfony and fetches messages over IMAP. Every email follows the same assessment path whichever way it arrived, so all of the chain's mail is sorted by one set of rules.

  2. 2

    A rules engine taught with the hotel staff

    We worked out the rules together with the hotel staff, who know their mail best. The engine reads the sender, domain, subject and body of each email. From these it tells bookings, date changes, group requests, invoices and spam apart, then sends the message to the right hotel and department.

  3. 3

    An AI layer for uncertain messages

    Not every email can be judged clearly by rules. Messages the rules are not sure about go to an AI layer built on OpenAI. The model decides whether the message is a booking, and the sorter uses that answer to choose where it goes. The AI only ever assesses a narrow slice of the mail.

  4. 4

    People only where a decision is needed

    Whatever neither the rules nor the AI can settle goes to a person. Reception no longer works through the whole inbox; staff only receive the messages that actually need a decision. The other 95% of the mail is split between hotels and departments automatically.

Outcome

Reception only sees emails that need a decision

The sorter routes 95% of the chain's mail automatically, with no reception involved. Bookings, date changes, group requests and invoices reach the right hotel and department. The rules were built on the hotel staff's knowledge, so the sorter splits the mail the way they used to. The AI layer helps only where the rules are not sure, and people see only what needs their judgement.

  • 95% of emails routed automatically
  • Rules built on the hotel staff's knowledge
  • AI used only for uncertain messages

We can take manual mail sorting off your team

We build Symfony automations that read, sort and forward messages by your own rules. Tell us how your team works with its inbox today and we will suggest an approach.