- United says personal AI assistants can show its flights but cannot complete a booking, citing privacy and regulatory concerns, according to The Verge.
- United worries agents could skirt federal rules about information airlines need to share with customers. The Verge's example is the explosives attestation.
- Leaders should map which customer steps need a person's consent or declaration, and decide how to treat agents there.
A checkout is more than a payment
A booking page looks like a shop counter. It is also where a traveler may be asked to confirm things, such as the explosives attestation the report cites. United has kept personal AI assistants away from that step. That shows where automated shopping meets human responsibility.
What United decided
United says it will not let personal AI assistants book tickets on its airline. The Verge reported this on November 11, citing The Wall Street Journal. United gave privacy and regulatory concerns as its reasons.
Agents can still show users United flights. They cannot complete the purchase for them.
The Journal says few people have tried to get an AI agent to book a United flight so far. The Verge says other businesses, such as Doordash, see the same.
How an agent works at a checkout
This section explains the general idea. It does not describe United's systems, which the sources do not detail.
A personal assistant that buys things acts through the user's own access. It may log in to an account and read pages. It may fill in forms and press the buttons a person would press. To do that, it uses the traveler's details, such as account access and personal information.
That creates the two risks United names. First, anything the agent can see or store could leak. Second, the agent works from loose instructions.
"Find me something cheap on Friday" leaves room for it to pick a ticket the traveler did not mean to buy. The sources do not say how often either problem happens.
The step an agent may not be able to take
United is concerned that agents could expose customers' information. It also worries they could make purchases customers did not intend. And it fears they could skirt federal rules about information airlines need to share with customers.
The report's example is having customers attest that their bags hold no explosives. The Verge's opening line says agents "can't attest that you don't have explosives in your bags."
The sources do not say who must make that attestation. They do not say what the federal rules require. This piece does not either.
Our view is that this is a human-accountability problem. An agent can fill in a form field. A person has to answer for what the field says.
The unintended-purchase worry is separate. The report does not say who would pay for a wrong ticket. Any business that sells things with real consequences should ask that question.
Not only United
The Verge says United is not alone in blocking unauthorized agents. It reports that eBay appeared to block agents from browsing its site. Several major banks have also placed restrictions on them.
These are separate decisions in different sectors. The report gives no count. They should not be read as a measured trend.
The word "unauthorized" matters. United says it could reconsider as the technology advances. In its words, "allowing them to interact directly with a United account or booking could involve sensitive actions, and we want to make sure we have established protections for travelers first."
The door is closed for now. The airline has not said what would open it.
The commercial side is plain too. CEO Scott Kirby said at an event on Tuesday, per the Journal: "To be honest, I don't want to create another middleman." An agent booking for a traveler would sit between airline and customer.
The lesson: showing is easy to delegate, signing may not be
This is our interpretation, not United's. Many sites assume a person is on the other end. That person can consent, declare and be held to what they said.
An agent breaks that assumption without breaking the page. The forms still load. The buttons still work. What may be missing is someone who can answer for the click.
That shifts the risk. Product teams will want agents to complete purchases. Legal and compliance teams will ask who made the declaration. For now, United lets agents look and keeps people in charge of the commitment.
Questions to put to your team
Start with a map. Which steps in your customer journeys carry a legal declaration, a consent or a payment authority? Those are the steps where an agent acting for a customer creates the most doubt.
Then ask what your policy is. Do you let agents browse, quote or buy, and who decided? Check what your logs show about agent traffic today, so the answer rests on data.
Finally, ask your own vendors the same thing. If a supplier's checkout needs a human statement, your staff's AI assistants may be refused there too. Better to learn that in a policy review than from a failed order.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: The Verge.
