I Let an AI Agent Book My Airport Parking
Airport parking is a small problem with an unusually large number of details.
The airport matters. So do the dates, the shuttle timetable, the side of the border, the price, the payment system, and the confirmation email. The task is simple in the way assembling flat-pack furniture is simple: every individual step is obvious, and the whole thing becomes annoying because there are so many of them.
For our trip to Iceland, I asked my AI agent to find the cheapest parking near Basel airport for 1–11 August 2026.
That sentence contained more work than it appeared to.
The first useful answer was a comparison
The agent looked at EuroAirport Basel Mulhouse Freiburg and compared the obvious options. Official airport parking was easier. It was also more expensive: the cheapest official option for the period was about 111 EUR.
Adopark, a nearby third-party provider in Bartenheim, was roughly 79–83 EUR, depending on the booking flow.
The important result was not the cheaper number. It was the condition attached to it. Adopark made sense if the shuttle worked for our schedule. Official parking made sense if I wanted to remove another moving part and was happy to pay for that convenience.
This is the point where search turns into assistance. A search engine can return two prices. An agent can put the decision between them into words.
A booking is a sequence of small decisions
I told the agent to book Adopark.
It opened the site and filled in the ordinary parts of the form: outdoor parking, the free airport shuttle, the dates, the passenger count, and the Basel airport transfer details.
Then it stopped.
The form required the car model, license plate, billing address, and phone number. The agent did not have enough information for those fields, and it asked me.
That pause was more reassuring than any smooth demo. A system that invents a license plate can complete a form. It cannot get a car through a car park gate.
I supplied the missing information. The order was ready, but one small decision remained.
Eighteen-thirty or nineteen hundred?
Our flight was at 21:20. The agent had chosen 18:30 as the drop-off time, leaving a generous margin for the shuttle and the terminal.
I thought this was too early and asked why.
Its reasoning was sensible. Eighteen-thirty was deliberately conservative; 19:00 would probably still leave us with about two hours at the terminal. Both answers were defensible. The difference was not knowledge. It was temperament.
I kept 18:30.
The calculation was easy. The preference was mine. If the purpose of delegating the task was to make the trip less stressful, saving thirty minutes was a poor bargain.
Sometimes the cheapest option is the one that leaves you with fewer things to think about.
The human returned at payment
The agent tried an agent payment flow first. It did not work for me in Germany, so it handed me the direct Payplug checkout instead.
I paid manually, sent the agent a screenshot, and it verified that the payment had succeeded.
This was a clean division of labour. The agent could inspect the options, carry the booking through the form, notice what was missing, and read the result. I could provide private details, make the final decision, and handle a payment flow that required my card and regional authentication.
I prefer that arrangement. The last click has consequences, and I want to know when I am making it.
Why this small task mattered
Nothing spectacular happened. There was no generated software, no elaborate demonstration, and no grand automation pipeline.
There was a parking reservation.
But the agent handled the parts that normally make these tasks irritating:
- it found the relevant trip context
- compared current prices
- explained the convenience-versus-cost tradeoff
- checked the shuttle constraint
- filled the booking form
- asked for information instead of making it up
- reconsidered the timing when challenged
- stopped at the sensitive payment boundary
- verified the result from a screenshot
This is a useful shape for a personal agent. It does not take over the decision. It reduces the amount of attention required to make the decision.
That distinction is easy to miss because the task is so ordinary. Yet ordinary tasks are where context switching accumulates. Fifteen minutes for parking, another fifteen for a receipt, another twenty for a renewal. None of them deserves your full concentration. Together they take an evening.
The future of agents may arrive this way: one dull edge at a time.
Airport parking has plenty of them.