A Closer Look at the Spread of Algorithmic Management Systems—Which Contributed to $8 Billion in Turnover Costs and a 45% Injury Rate at Amazon—Across Corporate America
With the help of digital technology, machines can perform complex managerial tasks, like supervising employees and assessing job candidates. It is no wonder that 57% of workers expect AI to change the way they do their current job, and 36% expect it to go as far as replacing their current job. With employees often described as a business’s most important assets and human capital a key factor in economic competitiveness, investors and policymakers alike must understand and manage the risks to businesses and workers stemming from the deployment of digital technologies. And because JUST Capital research shows that 51% of workers at Russell 1000 companies, who make up 15% of the employed population in the US in 2021, are not earning a living wage, defined as sufficient to afford a decent standard of living for the worker and his or her family in a particular place, long-term investors must also understand and mitigate the harms that AI can cause to low-wage workers.
While there has been a recent surge of interest in and funding for AI safety or mitigating catastrophic societal-scale threats of uncontrolled AI systems, ex/ante founder and managing partner Zoe Weinberg explains that more proximate harms—like discrimination, bias, and fairness—are taking back seat. The time has come for long-term investors to take the drivers’ seat to first understand and address these harms.
The ways in which AI can impact job quality are manifold. This article—the first in a series on AI and the workplace—focuses on algorithmic management, the delegation of managerial functions to algorithms in an organization, which is becoming an important part of AI-driven digital transformation of companies. This article closely examines the impact of algorithmic management on Amazon to illustrate why algorithmic management systems should matter across portfolios to all investors.
The Promise of Algorithmic Management Systems
Organizations increasingly rely on algorithmic-based HR decision-making to monitor employees, driven in part by the technology industry’s claims that its decision-making tools are efficient and objective. Algorithmic management systems can increase the scale of management operations: matching Uber’s 149 million riders with its over 7 million drivers in 10,000 cities across 70 countries through algorithms is indeed unprecedented. Companies also harness algorithms to increase efficiency: UPS trucks have long been equipped with computers that give drivers advice and filled with sensors that record when drivers open doors, buckle their seat belts, and back up their trucks. In addition, algorithmic management systems can make employees more effective: MetLife call center workers receive real-time feedback from AI on whether they are not empathetic enough, sound tired, or speak too quickly.
A Slippery Slope
However, researchers from MIT find that focusing solely on efficiency can lower employee satisfaction, wellbeing, and performance in the long-run by treating workers like “cogs in a machine” or triggering employees to continue working to the point of exhaustion.
As Interfaith Center on Corporate Responsibility (ICCR), a coalition of faith-based and values-based investors, and OpenMIC, a nonprofit focused on responsible use of digital technologies, explain in their new report, Dehumanization, Discrimination and Deskilling: The Impact of Digital Tech on Low-Wage Workers, a critical element of algorithmic management systems is the monitoring and surveillance of workers in violation of their human rights. More specifically, the report outlines a number of human rights impacted by digital technologies, including right to privacy (Article 12 in the 1948 Universal Declaration on Human Rights) and occupational safety and health (ILO Convention 155 in 1981 and 187 in 2006).
Why This Should Matter to Investors
Worker well-being, satisfaction, and human rights should matter to all investors. When algorithmic management systems detract from them, the resulting higher worker turnover, higher injury rates, increasing regulatory fines and sanctions, and increasing regulation can materially dampen long-term value creation.
Tying Worker Satisfaction and Safety to the Financial Statements at Amazon
Amazon—the second largest private employer in the US and the fourth largest US company by market capitalization—is worth closely examining because of the scale of the impact of algorithmic management on labor wellbeing and shareholder value. The lessons learned from this close examination are relevant to other companies that also use algorithmic management.
Over the past couple of years, researchers, investigators, lawmakers, and workers have been reinforcing each other’s claims that Amazon’s high injury rates are linked to the company’s workplace productivity quotas and surveillance practices. For example, Amazon warehouse worker Daniel Olayiwola, who became the first warehouse worker in corporate history to present his own resolution at Amazon’s annual shareholder meeting in 2022, highlighted the dangers of Amazon’s “Time Off Task” (TOT) policy, which tracks the amount of time per day to the minute that workers do not scan products (including during bathroom breaks), its rate system, the number of products that employees scan per hour, and algorithmic firing based on TOT and rates.
During the pandemic, the New York Times reported 3% turnover a week or 150% turnover a year, which means that on average Amazon had to replace its entire workforce every eight months. According to leaked internal documents marked “Amazon Confidential” that tech news website Engadget reviewed and reported on, the cost of attrition was an estimated $8 billion annually for Amazon in 2021—nearly 25% of Amazon’s 2021 net profit of $33 billion. Engadget further reported that regretted attrition, or workers choosing to leave the company, occurred twice as often as unregretted attrition, or layoffs and firings, and that only one out of three new hires in 2021 stayed with the company for 90 or more days. Amazon’s announcement last week to increase the average total compensation package for fulfillment and transportation employees in the US to over $29 per hour—Amazon’s largest ever investment of over $2.2 billion in pay and benefits—may be a cost-effective way to reduce the $8 billion cost of employee turnover or said differently, it may be hazard pay.
Because it’s challenging to update information gleaned from leaked confidential documents, and we must try to understand worker satisfaction and algorithmic monitoring following the pandemic, let’s also discuss the interim report released by the Senate Health, Education, Labor, and Pensions (HELP) Committee, which was investigating workplace safety.
According to the report, during Prime Day week in 2019, Amazon’s rate of recordable injuries—the injuries that Amazon is required to disclose to the Occupational Safety and Health Administration (OSHA)—was over 10 injuries per 100 workers. In addition, the report explains that Amazon’s total injury rate, which includes injuries that the company is not required to disclose to the Occupational Safety and Health Administration (OSHA), was nearly 45 injuries per 100 workers during the week of Prime Day in 2019. That is almost half of the company’s warehouse workers! The report highlighted that although Amazon’s total injury rate includes minor injuries, such as bruises and superficial cuts, it also includes serious injuries, such as torn rotator cuffs and concussions, which the company should have treated as recordable injuries.
Amazon examined the issue closely in a memo titled “2021 Prime Day Lessons Learned,” which analyzes staffing performance for the period leading up to Prime Day in 2021. As context, Amazon had 117,300 employees in 2013, growing steadily to the peak of 1.6 million in 2021 before declining in 2022 and 2023 to 1.5 million. The document states that Amazon had only a 54.7% success rate meeting its hiring target from the end of March through the beginning of May in 2021. That means Amazon filled barely more than half of the positions it needed to fill to have fully staffed warehouses during that period. From the beginning of May through the end of June—ending the week of Prime Day—the company met only 71.2%. Indeed, worker satisfaction and worker safety are deeply intertwined.
Tying Worker Satisfaction and Safety to Regulatory Risk at Amazon
Several states, including New York and California, passed bills that require companies to disclose production quotas to warehouse workers and prevent companies from retaliating against employees for failing to meet undocumented quotas. US Senator Edward Markey (D-Mass) earlier this month introduced the Warehouse Worker Protection Act at a federal level, which if passed would require companies to provide written notice to workers of quotas and prohibits dangerous quotas, including those that rely on constant surveillance.
In addition, the Stop Spying Bosses Act, introduced by US Senators Bob Casey (D-PA), Cory Booker (D-NJ), and Brian Schatz (D-HI), would require any employer collecting data on employees or applicants to disclose such information in a timely and public manner, prohibit employers from collecting sensitive data on individuals (i.e., off-duty data collection, data collection that interferes with organizing, etc.), create rules around the usage of automated decision systems to empower workers in employment decisions, and establish the Privacy and Technology Division at the Department of Labor to enforce and regulate workplace surveillance as novel technologies evolve and grow.
As a case in point, two warehouses violating California’s Warehouse Quotas law cost Amazon $5.9 million in June 2024 alone. If passed, the federal bills would expand the scope of costly fines. The biggest threat to Amazon is likely to be action by government regulatory bodies. Amazon’s largest regulatory fine was EUR 746 million (US$887 million) in 2021 for violating the General Data Protection Regulation (GDPR), a European privacy law. More recently, in January 2024, French regulator CNIL fined Amazon France Logistique EUR 32 million (US$35 million) for an “excessively intrusive” surveillance system: CNIL said that tracking the inactivity of employees’ scanners was illegal and that the system set up to measure the speed at which items were scanned as “excessive.” As US regulation catches up to EU regulation on AI and privacy, regulatory risk could weigh down Amazon stock price, which at $187.99 per share nears all-time highs.
Spreading Like a Virus
Amazon is not alone. Algorithmic management systems are pervasive. According to researchers at the UC Berkeley Labor Center, across the country, employers are increasingly using data and algorithms in ways that stand to have profound consequences for wages, working conditions, and equity.
Countless workplace surveillance systems collect data about worker activities through a wide variety of means, including handheld devices, point-of-sale systems, mobile phones, fingerprint scanners, fitness apps, wellness apps, smart cameras, microphones, and body sensors, according to Aiha Nguyen, who explores the future of labor for Data & Society, a nonprofit research organization. Nguyen says algorithmic management systems enable “work speedups, employment insecurity and instability, a shift of risks of and costs from employers to workers,” among other issues.
The AI-driven digital transformation of companies is only beginning, and it would behoove long-term investors to understand and shape it. This series of articles—which will also cover exploitation of data workers, the gig economy, and automated hiring tools—aims to do just that. Investors that can proactively identify and reduce exposure to these harms in their portfolios should be better able to generate attractive risk-adjusted returns and immunize their portfolios from regulatory risk.
The views expressed in this article are the author’s own and do not reflect the views of any of her affiliations.