Track every booking confirmation with one travel address

You know the ritual. Standing in the check-in queue, scrolling frantically through your inbox for a booking reference buried under three weeks of newsletters, receipts and calendar spam. Was it under 'Your e-ticket'? 'Booking confirmed'? The airline's name, or the travel agency's? Every airline formats its confirmation differently, and none of them care that you'll need that six-character code while holding a passport in your other hand.

The fix isn't a better search query. It's making sure travel mail never mixes with everything else in the first place. And, once it's isolated, letting software pull out the parts you actually need.

Create a dedicated address and use it for every airline account, hotel loyalty programme, booking site and train operator: travel@yournick.hidemy.world. On HideMy.world the address exists the moment you first use it, so there's nothing to set up before your next booking: just type it into the checkout form.

From that moment, the address is a perfect classifier. Anything arriving at travel@ is, by definition, travel. No sender lists to maintain, no filters that break when an airline rebrands its mailing domain or a booking site routes confirmations through a partner. And when a booking site inevitably starts sending 'deals we thought you'd love', your rules deal with it in one place rather than in every mailbox you own.

Travel addresses attract two kinds of mail: the confirmations, tickets and schedule changes you genuinely need, and an ocean of promotional filler. Telling them apart with keyword filters is miserable: 'confirmation' appears in marketing subject lines, and genuine schedule changes sometimes arrive with cheery subjects like 'An update about your trip'.

This is exactly the kind of fuzzy classification AI rules are built for. Instead of enumerating keywords, you describe the intent in plain language and let the model judge each email:

Here's where it gets satisfying. A confirmation email is unstructured prose wrapped around a handful of facts: a booking reference, a flight number, dates, airports. A transformer extracts those facts into clean JSON, and because the extraction is described by intent rather than by position in the HTML, one transformer handles wildly different templates.

Consider the two extremes. Air France sends a dense, multi-column HTML layout where the reference sits in a table cell labelled 'Réservation', sometimes in French even for English-language bookings. Emirates sends a long scrollable ticket where the PNR appears twice, once near the top and once inside the fare breakdown, formatted differently each time. A regex that catches one will miss the other, and both airlines redesign their emails often enough that positional parsing is a treadmill. An intent-based transformer sidesteps all of it. You specify the output shape once:

Structured data wants a destination. For travel, the best one is the device already in your hand at the airport: route matched confirmations to a Telegram channel, with a message template that surfaces the extracted fields directly in the notification.

Give travel its own chat rather than mixing it into an existing one, and consider a shared group if you travel with a partner or a team: everyone sees the reference, nobody plays inbox-forwarding tennis the night before a flight. Because only rule-matched mail reaches the chat, it stays pristine: no promotions, no noise, just bookings.

A template like this turns every confirmation into a glanceable card:

You book a flight. Thirty seconds later your phone buzzes: airline, flight number, reference, route. No opening an email app, no searching. At the airport, the reference is the most recent message in a Telegram chat that contains nothing but bookings, scrollable in chronological order like a travel diary. A schedule change lands the same way, which means you find out about the gate change from a notification rather than from a departure board.

Because the pipeline emits JSON, it also composes with anything else you fancy building. Point the same pipeline at a webhook and you can drop each trip into a spreadsheet, build a personal travel log, or auto-create calendar entries with the extracted times. The Telegram ping is the instant win; the structured data is the foundation.

The whole arrangement takes about ten minutes: create the travel address, write the AI rule describing what counts as a confirmation, pick or define the extraction schema, and connect Telegram. After that it's maintenance-free: the next booking you make flows through automatically.

The deeper lesson travels well beyond travel, too. Any category of email that follows a pattern (orders, invoices, appointment reminders) yields to the same three moves: isolate it with a dedicated address, classify it by intent, extract the facts. Your inbox was never a good database. Now it doesn't have to be one.