The problem
A small team was manually reading every inbound support email and forwarding it to the right person. It took two people most of a morning, and urgent issues sat unread behind newsletters and invoices.
What I built
Built a pipeline that pulls mail via the Gmail API, classifies each message by intent and urgency, extracts the customer and order references, and routes it to the right queue. An eval set of 500 hand-labelled emails runs on every change so accuracy is a number, not a feeling. Anything the model is unsure about goes to a human review queue rather than being guessed at.
The outcome
Two people got their mornings back. Urgent issues now surface within minutes. The review queue catches roughly 6% of messages, which is also where the next round of training data comes from.
Need something like this?
If this looks close to a problem you have, the fastest way to find out is a short call.