I sat through a product demo where the presenter promised agents that would handle bookings while humans slept. You could feel the pitch outrun the reality, like a theatre with the lights on and no audience. I kept asking the practical question: the bots are ready—but who’s actually using them?
United banned AI agents from making bookings after citing privacy and safety concerns.
I read United’s announcement and felt the corporate red flags—privacy exposure, third-party data sharing, and regulators breathing down the neck. The airline said AI agents can’t make bookings because bots might accept terms or confirm baggage statements that should be human attestations, possibly running afoul of FAA and Department of Transportation rules.
That’s a blunt authority cue: when airlines point to federal rules, you don’t treat the feature as an optional checkbox. You start asking who really holds responsibility when a machine clicks “confirm.”
Can AI agents book flights or hotels for me?
They can attempt to, but companies are pulling the brakes. United won’t accept bookings from agents; Airbnb’s Brian Chesky told investors that Muse and Instinct are delivering “very non-material” traffic and even fewer bookings. In practice, the legal and safety constraints make many travel use cases off-limits for autonomous intermediaries.
Uber and DoorDash logged near-zero agent-driven orders even at peak demand.
On the week of Halloween, Uber’s platforms handled 320 million trips and deliveries, yet the company disclosed that just 500 of those were placed by AI agents. DoorDash, which rolled out an agent-text ordering feature in September, reported “maybe several hundred orders in total” so far, according to CEO Tony Xu.
Numbers like that are a direct signal: the market test failed to find traction. These bots arrived with fancy promises but behaved like a Swiss Army knife without the blade—an attractive tool that doesn’t cut where it needs to.
Are AI agents replacing human customer service?
Not yet. You and I can feel the tension: vendors pitch automation as labor replacement, but the real-world metric is adoption. When customers don’t hand over their transactions, businesses won’t build systems that force the issue. That inertia protects jobs and forces builders to solve for trust, not just capability.
Airbnb, investors, and product teams keep complaining about negligible agent traffic.
Brian Chesky’s “very non-material” phrasing is investor-speak for “this isn’t changing our business.” For product teams that hoped agents would instantly scale bookings, that realization is a cold pivot. You can have a capable automation; what you can’t buy is demand.
There’s also a user-experience truth: many tasks these agents promise to handle already have extremely low friction. Ordering food or booking a ride often takes a few taps—adding an intermediary can inject delay, uncertainty, or privacy concerns that customers rightly refuse to accept.
Why are companies blocking AI agents from bookings?
Regulatory risk, customer distrust, and the practical exposure of sensitive data top the list. United explicitly cited the danger of bots checking boxes about baggage contents or accepting transactions a human wouldn’t. When regulators and airlines warn, companies respond by locking the doors.
I’ve spent years watching emergent tech get sold as inevitable. Here you can see the reverse: a technology with glossy demos and confident founders running into mundane human choices—privacy, convenience, and a refusal to hand over control. You can call it a market correction or a slow exhale, but either way the machines aren’t finding jobs to replace yet—so who will teach them to be wanted?
