Amazon Says It Will Retrain Workers It’s Automating Out of Jobs. But Does 'Upskilling' Even Work?

By Brian Merchant on at

Amazon made waves when it announced earlier this month that it planned to spend $700 million (£558 million) to retrain one-third of its US workforce—100,000 employees—as part of its ‘Upskilling 2025’ initiative. The sizeable commitment, which the Wall Street Journal notes is “among the biggest corporate retraining initiatives” yet announced, will fund pilot programmes, classes, and tuition fees for employees. “The American workforce is changing,” Amazon’s announcement proclaims, and “there’s a greater need for technical skills in the workplace than ever before.” Translation: We will be automating your job soon, and if you want to keep a decent paying gig here, study up.

As is often the case when the online retail giant makes a sweeping, headline-grabbing announcement, we’re left with at least as many questions as answers. The first among them being: Is upskilling—or retraining, or reskilling—that many employees even plausible? If it were, what would it look like in practice?

It’s a pretty important question to answer, because ‘retrain the workers’ is probably the most common policy idea we hear from politicians and business leaders whenever they’re confronted with questions about how we might cope with an increasingly automated economy. Every conference panel discussion, political talking point, and sobering news report about automated job loss seems to feature pablum about the urgent need to educate workers to prepare them for the modernising world.

And it’s safe to say upskilling is already being embraced among corporations: major American companies like AT&T, JPMorgan Chase, Accenture, and, now, Amazon have all made splashy investments in programmes that purport to future-proof workers. According to a survey by McKinsey, 66 per cent of executives see “addressing potential skills gaps related to automation/digitisation within their workforces as at least a top-ten priority.”

“I think it’s extraordinarily positive that Amazon is doing this, and I think there are things they can do to make sure it succeeds,” says Jane Oates, a former Obama administration official who helped modernise some of its federal retraining programmes and the president of WorkingNation, a nonprofit founded by venture capitalist Art Bilger to spread the word of the coming employment crisis.

Yet America’s largest private sector labour union, United Food and Commercial Workers (UFCW), sees Amazon’s retraining initiative as little more than a way for the company to gloss over the fact that it plans to eliminate its workers’ jobs. “Jeff Bezos’s vision is clear—he wants to automate every good job out of existence, regardless of whether it’s at Whole Foods, Amazon warehouses, or competing retail and grocery stores,” UFCW President Marc Perrone said in a statement.

“Jeff Bezos’s vision is clear—he wants to automate every good job out of existence, regardless of whether it’s at Whole Foods, Amazon warehouses, or competing retail and grocery stores.”

The Union Network International (UNI), which has been assisting Amazon workers around the world in their efforts to organise, points out the initiative was rolled out without the input or participation of workers. “Amazon’s history of disregarding the safety and rights of employees raises red flags about it unilaterally restructuring operations without the input from workers and their unions,” UNI General Secretary Christy Hoffman wrote me in a statement. “Until Amazon negotiates with workers, many of the same problems—unfair scheduling, anti-union harassment, and extremely pressurised working conditions—will remain, even if the nature of the work has shifted.”

As Perrone puts it, “Amazon is throwing money at a problem it created and somehow thinks that it deserves applause. … Amazon has become an economic arsonist that suddenly decided to put out the fires it is starting.”

But the question remains: Could Amazon put out this fire through mass retraining, even if it really wanted to? Can it feasibly re-educate a full third of its sprawling American workforce?

In the US, the history of corporate job retraining has been spotty to say the least. The WSJ recently ran a story headlined ‘Why Companies Are Failing at Reskilling,’ which detailed the myriad shortcomings in the arena, including the fact that many employees were unable to meaningfully access the programmes.

Federal job-retraining efforts have historically produced dismal results, too, sometimes fantastically so. One of the most comprehensive studies of federal programmes designed to retrain dislocated workers, published by the US Department of Labour in 2008, concluded that “the gains from participation are, at best, very modest, even three to four years after entry. Overall, it appears possible that ultimate gains from participation are small or nonexistent.”

That may be why Andrew Yang, the Democratic presidential candidate campaigning behind a promise to address the looming threat of mass job automation, often calls out retraining programmes. “The efficacy level of federally funded retraining programmes for manufacturing workers was approximately 0-15%,” he tweeted in March (without citing a source). “Almost half of the workers left the workforce. If retraining is the answer we would need to be much better at it.”

There were some bright spots in the Obama administration’s efforts to upgrade the long-ailing federal job retraining apparatus, and US initiatives have long suffered from a relative lack of funding and resources (out of 29 developed nations, the US is second to last in money spent). Yet the highest profile examples are those like Janesville, in the US state of Wisconsin, where GM closed a major manufacturing plant and served as the subject of Amy Goldstein’s haunting book about the impacts of mass job loss—and showed federal retraining to be helpless to revive the town that depended on those jobs.

According to the DOL report, other studies, and surveys of America’s failed experiments with retraining, there are a lot of reasons that reskilling programmes flounder. Older workers are reluctant to reenter the classroom environment. Employees are wary of investing time, energy, and resources on retraining if there isn’t a guarantee that a job will be waiting for them when they’re done. Employers usually make retraining voluntary, after-hours, and don’t pay employees to undertake it. (Amazon, for its part, says it will cover 95 per cent of tuition, and offers employees “paid study time” but doesn’t elaborate. The company did not respond to my request for comment.) And then employees face the challenges of entering a brand new field—and of being inexperienced in it.

“The challenge is you’re trying to take workers who were reasonably well paid, far into their career, and train them for a new type of job,” says Carolyn J. Heinrich, the lead investigator in the aforementioned DOL study and a professor of public policy at Vanderbilt University. “Workers come out and earn entry level wages. In no way do they earn what they were making in the past.” Mobility is an issue, too. “Some of the jobs they trained for will be in other states—if you’re paying off a mortgage on a house, or kids in high school, you’re not going to be able to move very easily.”

All of these are going to be problematic for any broader program that tries to address automation with retraining, but some are less likely to afflict companies like Amazon, where lower-skilled workers are not making such comfortable wages in the first place.

“What’s different here is you’ve got a company retraining its own workers for jobs it knows it’s going to be able to provide,” Heinrich says. She says that this entire initiative is probably borne out of a cost-benefit analysis that concluded the best way to fill currently empty and future roles in an extremely tight labour market was to invest in retraining its own workforce. Amazon is currently trying to fill 20,000 roles; retraining current employees who’ve proven amenable to Amazon for those and future jobs, at a cost of $7,000 (£5,578) per employee, could be cheaper than recruiting. “You have to look at the employers’ perspective—what’s going to make them profitable and successful? It’s much easier for Amazon to have data on who they think is most likely to be successful in training, instead of having to go out and find these people. It should be a cost savings for them to do that within their own workforce.”

Remember, Amazon is huge, and it keeps reams of data on its employees and their productivity. They may have reason to believe that a particularly efficient fulfilment centre associate would make for an effective, say, sales or IT worker.

“When we looked at the problems with the worker training programmes it was problems with soft skills—showing up on time, poor behaviour in the workplace,” Heinrich says. “Amazon can be selective with who they choose—how much and what kind of mix of skills a candidate’s going to offer—does someone have the basic facility they need to learn coding? Now, if they can’t learn coding they’ll be less likely to provide the training to get them to that point. Maybe they’ll have that data already.”

“When Amazon commits to larger scale upskilling, who is going to be chosen for those kinds of jobs?”

And that’s one problem Heinrich sees with this kind of retraining—it’s inevitably going to skew towards better educated, savvier workers.

“You almost see, these days, two tiers of workers in manufacturing,” she says, “in the sense that employers have some of the lowest wage workers not on full time, or they’re hiring from temp agencies to meet demands at different times—and those are workers doing the lowest wage jobs, and are the least ready even in their soft skills to take on this kind of training. There’s always going to be someone left behind, right?”

Jane Oates of WorkingNation says that the first and perhaps most important step is to ensure that the training is part of the workweek. “People are busy—making this as easy to get to as possible will determine its success.” Amazon should also ensure they cover the tuition in the programmes outright, she says. “A lot of the failure is due to the tuition assistance programmes—a lot of the reason they fail is you have to pay upfront for the university yourself and get reimbursed,” she said. “And then if you don’t get a grade you’re out that money.”

Oates is encouraged that Amazon is covering the vast majority of tuition and that they’re funding training for programmes that fall outside the purview of the company’s own remit, like nursing. “Hopefully they’re going to have discussions with their employees about what’s interesting,” she says. “Hopefully they’ll do some preliminary work so people can self-assess or take a small course to see—is this something I would enjoy doing before they invest in a large amount of time in it?”

“The best of these programmes guarantee people an interview,” she adds.

Again, as of now, Amazon’s announcement is short on details, so whether the programme will include any of those prescriptions remains to be seen. Just how seriously the company ends up taking this initiative at all is an open question, too—Amazon committed to running its operations on 100 per cent clean energy, after all, and stalled out at the halfway mark without offering up any explanation.

But this isn’t ultimately about the workers. As Heinrich noted, it’s a calculated investment in business operations. Amazon has said it will take about 10 years to automate its warehouses; other facilities and departments will be following suit, and Amazon is providing a pathway for its more ambitious (and indefatigable) employees to continue to help the company turn a profit in less traditional roles. Like its competitors, Amazon is presumably seeking to downsize its labour force and cut operating costs in the long run, and, whether through attrition or otherwise, job numbers are going to winnow. Amazon doesn’t need 300,000 IT workers.

That’s why the UCFW is up in arms about Amazon’s programme—it sees the writing on the wall. It’s why its president, Marc Perrone, ended his statement by saying, “It is time to realise that Amazon’s ruthless business model will lead to massive job losses that could cripple our entire economy.” Retraining may move a handful of capable warehouse employees into technical work, but by and large the net number of good jobs the company supports is still going to decline, probably steeply. It already has: Warehouse worker hiring plateaued, despite record sales, for the first time last year.

Combined with the dubious history of jobs retraining programmes, this all generally paints a grim portrait for the future of a large swath of current Amazon workers—and for the feasibility of America’s favourite automation solution. (This, by the way, does not mean refusing to invest in such programmes; better, smarter retraining programmes that actually support workers while they’re studying and give them more agency generally, as those run in places like Germany, France and Sweden, are more effective.)

Retraining will likely indeed work for some—but probably not many, and probably not those who need the jobs the most. Which is precisely why the only reasonably satisfactory path forward involves giving workers a seat at the table.

“When Amazon commits to larger scale upskilling,” Heinrich says, “who is going to be chosen for those kinds of jobs? If they want their employees to advance, how much are they going to invest in them?” Anyone familiar with Amazon’s record in dealing with labour issues at its facilities should already know the answer to that. “If you have a worker who’s not as well prepared, is that worker going to be passed by? That’s what’s happened in the past.”

Featured image: Julio Cortez (AP)