The travel technology industry has poured billions into automation. Search engines, booking flows, payment processing, e-ticketing — most of the routine travel booking process is now fully automated. And yet, the most valuable moments in travel still require a human being.
This isn't a failure of technology. It's a recognition that travel is one of the most complex consumer transactions that exists — involving real-time inventory, regulatory compliance, cross-border payments, and deeply personal preferences. Automation handles the 80%. Humans handle the 20% that matters most.
The Edge Cases That Break Automation
Every travel platform encounters them. They're the tickets that can't be booked through standard flows, the requests that don't fit neatly into dropdown menus, and the situations that require judgment — not just execution.
Infant and Child Bookings
Booking a flight for an infant (under 2, lap seat) sounds simple. In practice, it's a nightmare for automated systems. Different airlines have different policies: some require the infant to be added to an adult ticket, others issue a separate booking. Some charge 10% of the adult fare, others charge a flat fee. Many LCC websites require a phone call to add an infant after booking.
A human operator knows the specific workflow for each airline and can handle it in minutes. An automated system would need to maintain integrations with dozens of different infant booking flows — most of which aren't exposed via API.
Name Changes and Corrections
A passenger's name is misspelled on the ticket. On a legacy carrier, this might be fixable through the GDS with a simple name correction. On a low-cost carrier, it's a paid name change — often costing €50-160 — and it can only be done through the airline's website or call center. Some airlines only allow name changes within 24 hours of booking. Others don't allow them at all and require a cancellation and rebooking.
No API can navigate this matrix of rules reliably. A human concierge can.
Wheelchair and Special Assistance Requests
Passengers with reduced mobility need special assistance at the airport — wheelchair service, priority boarding, medical equipment clearance. These requests are typically submitted as SSR (Special Service Request) codes in the GDS, but the format and acceptance criteria vary by airline. For LCCs, special assistance must be requested through the airline's accessibility portal — a completely separate system.
Getting this wrong doesn't just create a bad experience. It can mean a passenger can't board their flight.
Visa Letters and Travel Documentation
Some travelers need visa support letters from their booking agent — confirming their itinerary, hotel reservations, and return flights for embassy applications. This requires generating custom documents that reference specific booking details, sometimes with notarization or company letterhead. It's a fundamentally human task that requires context, judgment, and sometimes negotiation with hotels and airlines.
LCC Booking: The Human Operator Requirement
As covered in our article on LCC coverage, low-cost carriers don't participate in GDS or NDC distribution. The only way to book them programmatically is through direct interaction with airline websites.
This means a real person navigating the airline's booking flow: entering passenger details, selecting ancillaries (baggage, seats, priority boarding), processing payment, and confirming the booking. RunRelay's trained operators handle thousands of these bookings daily, with an average completion time of 8 minutes per transaction.
The result for the API consumer is seamless. They receive a structured webhook with the PNR and e-ticket. But behind the scenes, a human made it happen.
Disruption Handling: Where Humans Shine
Travel disruptions — cancelled flights, delays, missed connections — are where the human-in-the-loop model proves its value most dramatically.
When a flight is cancelled, the automated response is typically a notification: "Your flight has been cancelled. Here's how to rebook online." But rebooking during mass disruption means:
- Evaluating alternatives across multiple airlines and routes — not just the original carrier
- Rebooking hotel stays that are affected by the changed itinerary
- Coordinating ground transfers when arrival airports change
- Handling insurance claims if delay compensation applies
- Communicating proactively with the traveler — often while they're standing in an airport terminal
The difference between a self-service platform and RunRelay during disruption isn't speed — it's resolution. We don't just notify you. We fix it.
RunRelay's concierge team monitors GDS schedule change feeds in real-time. When a disruption is detected, the team initiates rebooking before the customer even knows something went wrong. Average resolution time: 12 minutes from disruption detection to new confirmed itinerary.
The 80/20 Model: AI + Human Concierge
RunRelay doesn't position humans against automation. The two work together in a layered model:
- AI handles the 80%: search, pricing comparison, standard bookings on GDS/NDC carriers, payment processing, ticket issuance, itinerary delivery, and routine modifications.
- Humans handle the 20%: LCC bookings, edge-case requests, disruption rebooking, special assistance, complex multi-city itineraries, group bookings, and anything that requires judgment or negotiation.
This isn't a compromise. It's a design choice. By routing the right tasks to the right layer, RunRelay achieves both the efficiency of full automation and the reliability of human oversight.
For API consumers, this is entirely transparent. Every request goes through the same endpoint. Every response follows the same schema. Whether it was handled by automation or a human concierge, the result is the same: a confirmed, ticketed booking with full lifecycle management.
Why This Matters for Your Platform
If you're building a travel product for a bank, fintech, or enterprise, your customers expect premium service. They expect every flight to be bookable. They expect disruptions to be handled. They expect someone to pick up the phone when things go wrong.
Pure automation can't deliver this today. Pure human concierge doesn't scale. The hybrid model — AI for the routine, humans for the critical — is the only approach that delivers both.
That's what RunRelay provides: a single API that's powered by intelligent automation with human concierge built in. No fallback. No escalation queue. Just execution.