Kate Crawford is a professor at the University of Southern California at Annenberg, a senior principal researcher at Microsoft Research and author of “Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence.”

The vote by Amazon warehouse workers in Bessemer, Ala., to join the Retail, Wholesale and Department Store Union has been defeated. Legal challenges and union organizing will likely continue in the effort to create the first U.S. union at the world’s fifth-largest employer.

This battle has wide-ranging implications. It could be a harbinger of the future of employment more generally, as the harsh realities of algorithmic management move deeper into the everyday experience of work and more power is concentrated into the hands of corporate leaders.

I’ve spent time inside Amazon’s fulfillment center in Robbinsville, N.J., to see firsthand the experience of working in semiautomated environments. The scale of the space is overwhelming: a 1.2-million-square-foot factory for sorting objects to be delivered, a dizzying spectacle of contemporary logistics and standardization. Bright-orange Kiva robots glide across the concrete floors, programmed to find the most efficient path to deliver heavy shelving units to human workers, or “pickers.” (Amazon founder Jeff Bezos owns The Washington Post.)

Compared with the algorithmic ballet of the robots, the workers looked less serene under the gaze of surveillance cameras. Several people I saw wore support bandages from the strain of repeated movements, and they all experience the pressure of making the “picking rate” — the time allocated to sort items.

Algorithms set the pace of work, decide where products should go, and determine the best size and strength of cardboard boxes for shipping. Apparently without irony, that algorithm is called “the matrix.” These are the background processes that enforce Amazon’s goal of rapid delivery on everything from dishwashing detergent to yoga mats. Humans are there to complete the specific, fiddly tasks that robots cannot.

While this might sound different from typical white-collar office jobs, many U.S. employers are eager to emulate Amazon as much as they can, and increase automation and surveillance systems at work. One industry study indicated that the covid-19 pandemic has accelerated this trend significantly, with the majority of U.S. businesses now reporting plans for greater automation. While robots might not be coming to take all jobs, a more insidious shift is underway. Artificial intelligence systems are increasingly used to track, assess and rank workers — often without their knowledge. This, in turn, acts as a force multiplier for the asymmetries of power between bosses and employees.

With the shift away from offices during the pandemic, employers are turning to “bossware,” software that tracks employee productivity as they work remotely. One such system is Interguard, quietly placed on worker’s computers to create a continuous profile of how much time is spent in productive or unproductive modes. It can also take snapshots of an employee’s screen every five seconds and send reports if they appear to be searching for jobs elsewhere. Companies such as Prodoscore similarly use artificial intelligence to score the daily productivity of each worker, compare them with colleagues and issue alerts for “high-risk” employees. Prodoscore also claims to have seen a 600 percent increase in interested clients since the pandemic began. What makes this kind of scoring and surveillance possible is that so much of contemporary office life is already mediated through software: email, Slack messages, phone apps and video conferencing calls. Taken together, this creates a highly granular and invasive data portrait, and workers are rarely able to opt out.

The future of work under regimes of algorithmic surveillance is less science fiction and more a supercharged version of the dehumanizing factory techniques established in the early 20th century. From Henry Ford’s production lines to Frederick Winslow Taylor’s micromanagement of human bodies, there is a long history of approaches that prioritized centralized control over time and productivity, where workers’ health was secondary to improving the efficiencies of production.

Today, we are witnessing new refrains of an old theme. Decades of labor organizing achieved improvements in working conditions, including the eight-hour workday, the 40-hour workweek, overtime pay and the right to organize, but the pandemic has been eroding some of these significant victories. Further, there’s little evidence that the turn to micro-level surveillance and worker scoring produces any consistent benefits. The downsides are clear, from increased employee stress to a substantial loss of trust at work.

While AI tools can augment human capabilities, many bossware tools do the opposite: They increase demands for productivity and availability, collapse what remains of the boundaries between work and leisure time, and threaten autonomy, privacy and dignity. Amazon has led the way on some of the most extreme forms of algorithmic management, seeking to extract maximum labor from workers just as it extracts data from users of its vast computational network. Regardless of this week’s Amazon result, algorithmic management will continue intensifying across workplaces, which means more of us will need to resist systems that treat humans like robots.

Read more: