What 20 years of sourcing taught me about AI in travel

Everyone in travel is being pitched AI right now. Most of the pitches fail the same test a bad supplier contract fails: the headline number looks great and the total cost is hidden in the exceptions.

Sourcing thinking, applied to AI

When you negotiate hotel or air content, the rate is never the whole price. You price the failure modes: what happens on cancellation, who eats the debit memo, how fast disputes settle. Twenty years of that trains a reflex — always price the exception path.

Apply that reflex to AI in operations and the picture gets clearer:

An AI agent that handles 90% of a workflow hasn't removed the workflow. It has concentrated the hardest 10% on your most senior people, while the easy 90% — which was your training ground for juniors — disappears. Price that.

Accuracy that degrades silently is worse than a queue. A human backlog is visible and embarrassing. A model quietly mis-categorizing supplier invoices is invisible until reconciliation. Instrument before you automate.

The vendor's demo data is their best supplier. Ask what happens with your worst one — the airline that sends malformed files, the hotel chain that disputes everything.

Where it's actually working

The wins I've seen are narrow and boring, and that's a compliment: fare-rule extraction, invoice matching, first-draft supplier communications with human sign-off. Each has a visible error path and a human owner. None of them are "we replaced the ops team."

The travel companies that get this right won't be the ones that adopted AI fastest. They'll be the ones that adopted it the way a good sourcing team signs a contract — eyes on the exception clauses.

Next up: how we structure supplier scorecards for AI-era procurement. Subscribe on LinkedIn to catch it.