As AmazonThe company's contract driver, Stephen Normandin (Stephen Normandin), spent nearly four years delivering parcels in Phoenix. Then one day last year, he received an automatic email. The algorithm that tracked him concluded that he had not done his job well.
The 63-year-old veteran was stunned. He was fired by a machine.
Mr. Normandin said Amazon punished him because things beyond his control prevented him from completing deliveries, such as locked apartment buildings. He said he was sad to be fired and proud of his strong work ethic. He recalled that when he was a soldier, he helped cook for 250000 refugees in Fort Chaffey, Arkansas, in (Fort Chaffee).
"I'm an old-school person, and I put 110% of my effort into every job," he said. "it really makes me sad because we're talking about my reputation. They said I didn't work, but I knew I did. "
Normandin's experience seems to be proof of decades of predictions that robots will replace workers. At Amazon, the machine is usually the boss-hiring, evaluating and firing tens of thousands of people with almost no supervision.
For years, Amazon has used algorithms to manage millions of third-party merchants in its online market, sparking complaints that sellers were wrongly accused of selling fakes and inflating prices and being driven off the platform.
The company is also increasingly ceding human resources operations to machines, using software not only to manage workers in warehouses, but also to monitor the performance of contract drivers, independent delivery companies and even office workers. Amazon CEO Bezos believes that machines make decisions faster and more accurately than people, which reduces costs and gives Amazon a competitive advantage, according to people familiar with the strategy.
According to interviews with 15 Amazon contract drivers who were fired, four of them said they had been mistakenly fired, and some former Amazon managers said the largely automated system could not fully adapt to the practical challenges that drivers face every day. The former executives say Amazon knows that delegating work to machines can lead to wrong and damaging headlines, but they believe that as long as drivers can be easily replaced, believing the algorithm is cheaper than paying people to investigate wrongful dismissals.
When drivers question poor scores, they don't know if they are communicating with real people. Responses usually contain only names or no names at all, and these responses usually apply to a variety of situations, not specific questions. Even with a name attached, the first few e-mail responses are likely to have been generated by a machine, according to people familiar with the matter.
When human managers are finally involved, they usually make hasty assessments-if they really want to do so-because they have to meet their own performance standards. A former employee of the call center supported by a driver said dozens of part-time seasonal workers with little training were sent to supervise the problems of millions of drivers
"Amazon doesn't care," said the former Amazon employee. "they know that most people will get the package, and those who don't get 2% or 3% will get it eventually."
Amazon's human resources operations are more automated than most companies. But it is becoming more and more common to use algorithms to make decisions that affect people's lives. Machines can approve loan applications and even decide whether someone should be paroled or put in prison.
After receiving Amazon's dismissal notice, Normanding began a familiar process, and as anyone could have encountered, he found himself in an automatic customer service cycle, but this time, Normanding is not looking for a refund for damaged products. He's trying to get his job back.
Within 10 days of Amazon's appeal, he sent a supportive email to the system asking for his dismissal to be withdrawn. He explained that he had reported to Amazon what was beyond his control and had been promised that he would not go against him because of these violations.
The next day, Normandin received a reply from Pavani G, thanking him for "providing more information about the historical background of your use of Amazon Flex". Normandin replied to the email, provided more information and received exactly the same reply, promising to investigate the issue, but this time it was signed "Bitan Banerjee". The email promised a reply within six days. Seven days later, a new email apologized for the delay and promised to update it as soon as possible.
At the same time, Normandin had no income. He expects Amazon's annual "Prime Day" promotion, which was postponed until October last year, to earn him the money he needs to pay his bills. On Oct. 19, Normandin sent a message to Amazon again, this time copying it to Bezos.
About 12 hours later, he received a reply informing him that Bezos had received an email and instructing Thai Taylor F to study the issue and respond on his behalf. On Oct. 23, Normandin received an email from Raquel on Amazon's Flex support team telling him that they were still reviewing his appeal. Former Amazon employees who used to work at Flex say escalating complaints to Bezos is a common strategy for drivers who have been stopped working, but it rarely helps them.
The final verdict arrived on October 28, and the email did not respond directly to Normanding's comments, but acknowledged the challenge of the work, saying: "We understand that every delivery partner has a difficult day, and you may sometimes encounter delays." we've taken that into account. " But Normandin still hasn't returned to his job.
He then tried several other delivery companies and finally decided to use his epidemic stimulus funds to start a small engine maintenance company. It's time to deal directly with humans again. For those who design algorithms to track, evaluate and eventually fire him, Normanding says: "they don't seem to have any common sense of how the real world works."