Automatic for the Bosses

Workers may be more affected by robots taking their bosses’ jobs than their own

Full-text audio version of this essay.

Generally speaking, the four economic sectors in the U.S. that rely most heavily on human labor are, in order of most people employed: retail, fast food, health care, and clerical office work. These are jobs that involve interacting — often intimately — with other people. To eliminate these jobs, companies couldn’t just replace humans’ role in production with machines, as they might on an assembly line. It would also require reducing their customers’ expectations for intimacy and human contact, and acclimating people to serving themselves.

So far, this has meant enculturating a willingness to use the self-checkout lane or talk to a computer on the phone, or, at the cutting edge of retail automation, navigate a store with no employees whatsoever. “Our checkout-free shopping experience,” Amazon Go’s website proclaims, “is made possible by the same types of technologies used in self-driving cars: computer vision, sensor fusion, and deep learning. Just Walk Out Technology automatically detects when products are taken from or returned to the shelves and keeps track of them in a virtual cart. When you’re done shopping, you can just leave the store.” If this technology were adopted broadly, it would have massive implications for culture, labor, and theft — the original Just Walk Out Technology.

Silicon Valley has never shied away from calling its products magical. They can make an entire army of workers disappear behind smoke and mirrors

But perhaps nothing has changed consumer expectations of service as much as the suite of apps that promise convenient “seamless” commerce. Hailing a ride, getting groceries delivered to your door, or accessing a city’s worth of spare bedrooms is easier than ever. In the dark ages, ordering a pizza meant talking on the phone and having cash on hand to give to another living, breathing person. Now with a few taps on a phone, the pizza can be ordered and paid for. And while video, record, or bookstores have (or had) their own irreducible charm, literally nothing beats Netflix, Spotify, and Amazon when it comes to efficiently moving product.

Silicon Valley has never shied away from calling its products magical, and in a sense, they are. They can make an entire army of workers disappear behind the smoke and mirrors of the user interface. Despite the immense computing power and network infrastructure necessary to process payments and route deliveries, an UberEats order still shows up as a ticket in a hot kitchen, a person makes it with their own two hands before boxing it up and giving it to another poorly paid worker who huffs and puffs up the three flights of stairs to deliver it to you. The app reduces the consumer’s interaction with workers and thereby discourages their solidarity. Someone’s convenience is another person’s drudgery.

What makes this convenience possible is a kind of automation, but not the version that replaces workers with robots. Rather, it is powered by cheap contract labor managed by sophisticated suites of software, sensors, and algorithms, a model that is branching out to all sectors of the economy. The transformation goes like this: Someone finds friction in daily life and smooths it over with underpaid labor, venture capital, and a slick user interface. In the early honeymoon phase, the company and its app gets adoring media attention that re-emphasizes the supposed bargain of consumer convenience for workers’ exploitation. The antagonistic relation between consumer and service worker is also built directly into the apps themselves, by rating systems that encourage and even reward users to pass judgment on workers. How many stars does this Somali immigrant deserve for driving you to the airport? What kind of review does this 60-year-old fry cook get? Consumers are lured by convenience and invited to tattle on workers if they are insufficiently subservient. Their self-centered expectations about customer service become a direct means for undermining workers.

Even though the pandemic has intensified use of services like Instacart, their workers have seen little benefit. While these companies make bigger profits, the cost of labor is offloaded to unreliable tips. Grocery store, warehouse, and food delivery workers are hailed as “essential” and celebrated with Blue Angels flyovers, but the praise feels more like charity and PR. In reality, they work under life-threatening conditions because they lack the leverage to secure something better. Norms that empower insulated consumers at the expense of workers who are taken for granted, if not rendered relatively invisible, have played a role in this.

But lockdown has forced many more workers to the other side of the app. Before the pandemic, many of those now working from home may not have given much thought to what it is like to be managed like this themselves, having an app snitch on them the way the “independent contractors” for Lyft or Fiverr are used to. They are likely unused to being on the other side of the convenience trade-off. They haven’t faced the possibility of being “deactivated” from their jobs when their ratings dip below 4.6 out of 5 stars, as Uber drivers do. They perhaps haven’t faced an “opaque algorithm for rating employees,” like the Instacart employees who described this as one of their grievances when they recently unionized in Skokie, Illinois.

Working from home is becoming an opportunity for employers to bring employees’ work behavior under more minute scrutiny, even while they are distanced from managing them, both physically and conceptually. The economic uncertainty and massive unemployment that has come in the wake of the pandemic gives employees less leverage over working conditions and sets the stage for ever more invasive surveillance. Many of the companies that are helping administrate remote working — including Zoom, Microsoft, Cisco, and Slack — are marketing their tools to bosses (who pay the licensing fees) and not employees, who are forced to use them. These systems allow for a kind of automated management, in which bosses are insulated from their workers by interfaces that do the dirty work for them.

Companies are already able to look in on workers’ video feeds and read private messages on services and devices the company pays for. Zoom, for a time, allowed bosses to track workers’ faces to make sure they are paying attention in meetings. An early Slack feature was staging competitions among employees to see who got the most reactions to their posts. Teams (Microsoft’s Slack competitor) also allows what bosses openly call on the company’s support site “snoop in” functionality. Outside these apps is even more surveillance. If you install a Mobile Device Management (MDM) profile on your phone, which is often required by employers if you want to read work-related email on your personal smartphone, your boss can get reports on your location, calls, text messages, and browsing history. Companies like Hexnode advertise that their MDM systems compile this information to help managers determine the productivity of their underlings. Setting your own hours and being your own boss — if working from home ever felt like that way at all, which for most it doesn’t — becomes a tortured maze of metrics and profiles, much as it is for the independent contractors of the gig economy.

It is only a matter of time before Slack and its competitors go from being tattle-tales to being more like HR Robocops

These sorts of tools facilitate the ability to uncover underperforming or, more nefariously, unionizing employees. Not long after Slack was seized upon by workers as an efficient tool for union organizing, the National Labor Review Board ruled that employees “do not have a statutory right to use employers’ email and other information-technology resources to engage in non-work-related communications,” which was widely understood to mean that you can be fired for trying to organize your workplace on a company Slack or email platform. It means that if your boss wants to read your DMs, they can request that information from Slack and not, say, a neutral arbitrator. What should be subjected to due process is rendered a matter of corporate customer service.

Slack is likely to become more and more embedded in the working day: Automated tools like the Now Virtual Agent can be implemented to help managers organize vacation time or route workers’ IT requests. It is only a matter of time before Slack as well as its competitors go from being tattle-tales to being more like HR Robocops, automating everything from hiring and firing to regular performance reviews and workforce reports, much as cloud services like Pipefy, Frevvo, and BambooHR already promise.

Labor-law scholar Ifeoma Ajunwa has shown how automated recruiting and résumé-sorting tools have been discriminatory along ageist, ableist, sexist, and racist lines. One can readily imagine the same biases being embedded in other forms of automated worker surveillance as they begin to play a larger role in labor relations. Amazon has already started using algorithms to help determine which of their Whole Foods stores are in danger of becoming unionized. The predictive power of algorithms — which are often presumptively taken as fair and objective — could be used to plan automated mass layoffs before the first union card is even signed, using factors that correlate with union support to fire employees without running afoul of labor laws, which prohibit firing someone solely for organizing.

If workers already must contend with employers’ obvious privacy intrusions — including keyloggers, periodic browser history checks, and access to email inboxes — how quickly will their spying within our home workplaces be normalized? Will we gladly take an expensive, ergonomic chair that also tells your far-away boss when you are not at your desk? Will all your smart home devices — the smart refrigerator, the Ring doorbell, or your Alexa — keep track of not only your own work but the relationship between workers? Through facial recognition attached to the camera outside your front door, the voice recognition in, your Alexa, and the MDM software on your devices, your boss would know everything about who was at the organizing meeting you hosted at your house. Then they will fire you on the premise that your smart fridge colluded with your bank to tell your insurance company that all that fast food and beer is putting you at risk for diabetes and they marked you as a future cost liability.

Such privacy concerns are likely here to stay, particularly for tech workers: Jack Dorsey announced that his two companies — Twitter and the payment-processing service Square — are developing permanent work-from-home policies. Facebook also recently announced that all its open positions in the U.S. are now “available for remote recruiting and hiring” and that “many of Facebook’s 48,000 employees around the world will be able to request a switch to remote work.” Mark Zuckerberg admitted that Covid-19 has only sped up an inevitable timeline: “If you’re long on VR and AR and video chat [like Facebook is], you have to believe in some capacity that you’re helping people be able to do whatever they want from wherever they are … this is a direction that I think we’re going to want to go in more anyway.”

Facebook is, to use the parlance of the tech industry, “dogfooding” a new generation of remote-work technologies — that is, it is having its own employees test experimental systems that will eventually be sold to other companies for employee surveillance. This is not a change in many tech companies’ missions so much as an extension of their ongoing efforts to reorganize industries for more efficient capital accumulation. Of the 10 million jobs added to the economy under Barack Obama, 94 percent of them, according to economists from Harvard and Princeton, were gig economy jobs. If, in the 2010s, Silicon Valley went after cabs, food delivery, and movie rentals, this decade will see Zoom, Microsoft, Google, Cisco, Slack, and Facebook reorganize office work and education to make it leaner, cheaper, and easier to control at a distance.

In Forces of Production, David F. Noble shows how industrial automation was never about making assembly lines more efficient by replacing humans with robots. The goal, rather, was to consolidate power in the hands of middle and upper management whose training in mechanical engineering and business administration would have more direct control over the production process. Working-class machinists, operators, secretaries, and assembly-line workers could do the job faster and better than the machines, but that was never the point. Indeed, production rates and quality of finished products would go down when new automation systems were first implemented, but the long-term benefit of eliminating not just labor costs but labor power was well worth it. One could view the imposition of Slack or Slack-like systems in similar terms: The point is not to make workers more productive; it is to make them more docile.

Automation has not merely been about reshaping work processes but creating conditions in which workers have less leverage — eliminating the time and space for solidarity to form

One of the more difficult parts of automation is capturing in text or code what Shoshanna Zuboff in In the Age of Smart Machines describes as the “oral cultures” within companies — the knowledge passed down among employees about how work processes and routines actually work. In modern workplaces, Zuboff writes, “the degree of orality that surrounds a set of practices is also related to the degree of power and autonomy that can be enjoyed by its practitioners.” This is in part a defensive measure. If a job can only be learned through slow enculturation, then bosses have a harder time replacing workers.

Automation is usually thought of as synonymous with deskilling, which makes it easier for employers to hire and fire workers made more interchangeable and thus cheaper. But deskilling faces a counterforce in credentialing, which regulates labor pools through requirements for prerequisites and certifications. As Harry Braverman concludes in Labor and Monopoly Capital, compulsory education has less to do with learning than with keeping “unemployment within reasonable bounds.” Secondary schools (and, one could easily argue though Braverman does not, college) are nothing more than “teen-sitting organizations” that calibrate the number of young people in labor markets, dolling out certificates and degrees that manage which portions of the population are allowed to compete for better-paying jobs. Deskilling and credentialing are not so much about how hard a job is to master than who is allowed to perform it.

This makes the idea of a “skills wallet,” touted by the Liberal Democrats in the UK, all the more dystopian. The proposal would give every British citizen money (with employer contributions) for lifetime continuing education, which sounds promising enough until you consider that its likely effect would be to introduce more layers of gatekeeping and administrative contact within the realm of employment. It’s not as though citizens would be learning to fish in the afternoon and read Francis Fukuyama in the evening; they would be encountering checkpoint after checkpoint meant to evaluate whether they have devoted more of their lives to learning “skills” that serve the interests of capital. If you cannot, for example, pass a class that teaches you how to operate some Facebook-developed work-at-home surveillance suite, then you might be left uncredentialed for thousands of jobs.

From the earliest keyloggers attached to stenographer’s typewriters in the 20th century to whatever Mark Zuckerberg is cooking up next, the goal has always been to rationalize and automate the entire value chain. But this has not merely been about reshaping work processes but creating conditions in which workers have less leverage. Once, this was more a matter of restricting or rescinding workers’ special knowledge of their work; now it is becoming a matter of automating management practices to eliminate the time and space for solidarity to form.

As staff positions and layers of middle management are eliminated, there will be a race to transfer as many parts of the management system to uncaring machines as possible. They will have no sympathy for the hustle; they will cut no slack. The plummeting costs of sensor packages and detection systems coupled with the aggressive rollout of 5G towers means the surveillance systems may become more extensive and aggressive, connected to physical barriers or automated transportation systems. Such infrastructures make it possible to automate the most extreme forms of managerial control: The factory lockout and the tracking of workers through the public sphere: at protests, meetings, or even just the wrong neighborhoods. New incentive schemes will also be on offer: The whole city can become the corporate campus with special access to gyms, bars, and hotels that would have previously needed to be in physical proximity to be effectively guarded and managed. Your bot boss will be everywhere and nowhere, as you navigate an opaque algorithm of corporate accountability.

In a 2016 Baffler article about the surveillance dangers of Slack, Jacob Silverman finds a 1987 report by the (since disbanded) federal Office of Technology Assessment titled, “The Electronic Supervisor: New Technology, New Tensions.” Silverman writes, “Workers of the early Information Age were right to wonder if computing would redefine what the OTA report called the ‘basic tension between an employer’s right to control or manage the work process and an employee’s right to autonomy, dignity, and privacy.’” They saw more than thirty years ago what many of us are only just now realizing: that firms desire control over all components of work, and to the extent that it is cost-effective and legal to spy on workers, they will.

That tension has reached a breaking point. Bosses now have the means and motivation to mediate the entirety of work, leaving room for bots and AI to watch, listen, and govern employees. Performance reviews can take on the character and nuance of an Uber ride rating with statistics about how often you speak up in meetings, how chipper you are in the morning, and how long it takes you to eat lunch. Even as the control over the when and where of work loosens, the character, speed, and affect of that work will continue to be brought under increased mechanistic scrutiny. In an increasingly automated world, autonomy, dignity, and privacy are not the things that scale.

David A. Banks writes about cities, technology, and society from Troy, NY.