Amazon’s latest delivery innovation has made a splash. Unfortunately, it was in a customer’s swimming pool.
A Texas woman, Lindsey Austen, was waiting excitedly for her first Prime Air order when an Amazon drone hovered above her backyard, opened its compartment, and dropped the parcel straight into the water.
It is an undeniably funny video. It is also a nice example of how quickly a futuristic customer experience can turn into a very public CX problem.
Lindsey Austen, who lives in Richmond, Texas, had gotten an alert that her Amazon order would be delivered by drone to her home for the first time.
Footage shows the drone hovering above her pool, then releasing her package into the water pic.twitter.com/yTtpUqv5XP
— Sky News (@SkyNews) August 20, 2026
In its official announcement on expanding Prime Air to nearly 500 cities and towns by the end of 2026, Amazon stated:
“Prime Air is built to operate responsibly, and the safety of the employees, customers, and communities Amazon serves is the top priority.”
In this case, nobody was hurt, and the drone did not crash. Still, a waterlogged package is unlikely to be the frictionless, under-an-hour delivery experience Amazon had in mind.
The incident raises a familiar question for enterprises rolling out customer service technology: what happens when the automation gets it wrong at the exact point the customer sees it?
The Order That Got Lost in Translation
Amazon is hardly alone in learning that automated experiences can unravel in highly visible ways.
McDonald’s faced a similar challenge at the very start of the customer journey.
In June 2024, the fast-food giant ended its automated order-taking partnership with IBM after testing AI drive-thru technology in more than 100 US restaurants.
The system was intended to make ordering quicker and easier. Instead, viral videos showed it adding unwanted items, mixing up orders, and, in one case, generating more than $250 worth of McNuggets meals.
McDonald’s had previously said the technology was accurate around 85% of the time. Yet that remaining 15% can feel rather more significant when a customer is staring at a drive-thru screen full of food they never ordered.
In a statement made at the time, McDonald’s said:
“The goal of the test was to determine if an automated voice ordering solution could simplify operations for crew and create a faster, improved experience for our fans.”
The company has not abandoned the idea of AI-powered drive-thru ordering altogether. Indeed, it said that “a voice-ordering solution for drive-thru will be part of our restaurants’ future.”
But the decision shows that customers are not always willing to absorb the growing pains of a technology pilot, particularly when they are hungry and in a hurry.
When the Bot Knows Better Than the Brand
If McDonald’s demonstrated the risk of AI misunderstanding an order, Air Canada showed the danger of a chatbot confidently misunderstanding company policy.
In 2024, the airline was ordered to compensate Jake Moffatt after its chatbot incorrectly told him he could claim a bereavement fare refund after purchasing a full-price ticket. When Moffatt later attempted to claim the discount, Air Canada refused.
The airline argued that the chatbot had provided the wrong information and that Moffatt should have checked the policy page linked in its response. The British Columbia Civil Resolution Tribunal was unimpressed.
“It should be obvious to Air Canada that it is responsible for all the information on its website,” tribunal member Christopher Rivers wrote. “It makes no difference whether the information comes from a static page or a chatbot.”
That is a line every CX leader considering generative AI should keep close by. A chatbot may be automated, but customers do not see it as a separate entity. They see the brand.
Klarna Learns That Speed Has Limits
Klarna’s experience offers another warning, this time from the post-purchase support stage.
The buy now, pay later provider initially claimed its AI assistant could perform the work of 700 full-time agents.
Yet, in 2025, CEO Sebastian Siemiatkowski admitted that the company’s focus on cost had led to “lower quality” and announced plans to strengthen human support:
“Really investing in the quality of the human support is the way of the future for us.”
That does not mean AI customer service has failed. Klarna still sees a major role for automation, as do McDonald’s and Amazon. But each example exposes the same weakness: technology is often judged by how smoothly it handles the happy path, while customers remember what happens when they fall off it.
That is where graceful failure paths matter. A customer whose drone delivery lands in a pool should not have to navigate a chatbot, upload three photos, quote an order number, and then wait several days for a human response (this is an example, not what has actually happened in this specific instance.)
The recovery should be quick, obvious, and owned by the brand: replace the item, apologize, and make it right before the customer’s video does the rounds online.
Occasional mishaps are inevitable as enterprises experiment with new technology. Customers can forgive a mistaken order, a confused chatbot, or even a parcel taking an unexpected swim.
What they will not forgive is being left to clean up the mess themselves, only to be told the experience is “industry-leading.”
