LinkedIn·Friday, 31 July 2026·27d ago

𝗧𝗿𝘂𝘀𝘁 𝗶𝘀 𝗮 𝗱𝗶𝗮𝗹, 𝗻𝗼𝘁 𝗮 𝘀𝘄𝗶𝘁𝗰𝗵. 𝗔𝗻𝗱 𝗺𝗼𝘀𝘁 𝗔𝗜 𝗿𝗼𝗹𝗹𝗼𝘂𝘁𝘀 𝗴𝗲𝘁 𝘁𝗵𝗶𝘀 𝗯𝗮𝗰𝗸𝘄𝗮𝗿𝗱𝘀. Two camps…

Joris Vanherp

Co-founder @ Maya - Helping travel companies captivate travelers and sell more trips

𝗧𝗿𝘂𝘀𝘁 𝗶𝘀 𝗮 𝗱𝗶𝗮𝗹, 𝗻𝗼𝘁 𝗮 𝘀𝘄𝗶𝘁𝗰𝗵. 𝗔𝗻𝗱 𝗺𝗼𝘀𝘁 𝗔𝗜 𝗿𝗼𝗹𝗹𝗼𝘂𝘁𝘀 𝗴𝗲𝘁 𝘁𝗵𝗶𝘀 𝗯𝗮𝗰𝗸𝘄𝗮𝗿𝗱𝘀. Two camps dominate travel right now, and both are wrong. One says deploy AI everywhere today. The other says AI can't be trusted with anything real. But when the stakes are a €5,000 family holiday, nobody wants ideology. They want evidence. A Sabre exec put it well in a recent Skift report: the mistake is thinking trust is a switch you flip. It's a maturity curve, and the progression matters. You start with augmentation, where AI drafts, summarises, and suggests while humans decide. You move to supervised automation, where AI executes narrow actions with explicit approval gates. Only then do you earn managed autonomy, where AI acts end to end, but only where reliability is proven. You don't earn autonomy by declaring it. You earn it with evidence. The key design principle isn't "keep humans in the loop." It's design for reversibility. Every capability needs a human fallback that's automatic, fast, and boring. That means running both in parallel for longer than feels comfortable: let the AI handle the interaction, let humans keep authority for the judgment calls, then expand autonomy only when the data says you should. That's the step most companies skip. They want the autonomy before they've built the fallback, which is exactly the wrong order. At Maya, we see the same pattern with the teams we work with: the best results come when they treat trust like an operational system, not a feeling.