AI and the Future of Work

Workers’ struggles will determine how the latest round of automation will affect labor.

By Robert Ovetz | November/December 2023

This article is from Dollars & Sense: Real World Economics, available at

issue 369 cover

This article is from the
November/December 2023 issue.

Subscribe Now

at a 30% discount.

The 149-day-long writers Guild of America (WGA) strike by 11,500 Hollywood screenwriters is one of the most important strikes in decades because it tamed the Hollywood corporate beast. Along with the 160,000 members of SAG-AFTRA, the actors’ union still on strike after nearly three months, the writers’ strike was perhaps the largest strike explicitly directed at the threat of artificial intelligence (AI) to human workers. After decades of disturbing dystopian film plots about humans being wiped out by AI and robots being smarter than us, the very people who wrote and acted in them saw AI coming for their own jobs. While such films once helped us work through our anxieties with this technology, today those who make them are living it. The strikes began only months after the high-profile release of Open AI’s ChatGPT reignited a global conversation about the impact of artificial intelligence on work. Along with other publicly available AI, such as the graphics generator Dall-E owned by the same company, AI now appears poised to do many kinds of work once thought exclusive to humans.

These developments are prompting an anxious discussion about the future of labor. AI is already being integrated into a wide range of jobs, including making hamburgers, caring for the sick, writing papers, doing complex math and computer coding, making illustrations, analyzing complex legal documents, and even deciding whom to hire and fire. There is also a growing list of examples of AI replacing human workers entirely. This past spring, the nonprofit National Eating Disorders Association replaced its unionized workers with its “Tessa” chatbot, though the group later removed Tessa after it was found to be giving dangerous advice to clients.

Some of the technology does not yet live up to the hype, but will it one day? Whether AI makes human workers obsolete, makes our work easier, or makes it even more exploitative will depend on the balance of power between the capitalist class and the working class. That struggle will not be fought out in the halls of government, where the AI corporations are already engaged in a charm offensive—such as during Senate hearings on AI this past May—in order to forestall a possible ban in favor of regulation.

Strong unions and an organized working class can dictate the terms of a new deal that redistributes the fruits of increased productivity or does away with capitalism and work as we know it. Our future is being decided now whether we are ready or not.

The WGA and SAG-AFTRA strikes clearly illustrate that although AI is unlikely to replace the entire human labor force, it is already making some kinds of work obsolete and deskilling others. Like all forms of technology before it, from the loom to the personal computer, AI is a terrain of class struggle. How that struggle turns out is yet to be decided.

André Gorz’s Two Paths

The threat of AI to human labor is hardly a new concern to scholars of work and capitalism. Nearly 40 years ago, French socialist philosopher André Gorz’s book Paths to Paradise foresaw the AI apocalypse and warned of two possible outcomes. In one possible future, productivity-increasing technology like AI would neuter working-class struggle by segmenting the working class into a small number of “elite waged workers” alongside a superfluous “mass of unemployed and precarious casual workers.”

Gorz’s other possible future celebrated the abolition of work and the redistribution of wealth. This is a possibility because, Gorz wrote, “in the fully automated factory, the quantity of living labor drops towards zero, and so does purchasing power distributed as wages. Automation abolishes workers: equally, it abolishes potential buyers.” Without buyers and workers, Gorz foresaw a world in which work was abolished and replaced by democratic control over the economy and a society moving us past capitalism.

But Gorz wasn’t very optimistic about the socialist possibility. He thought the dystopian future of automated capitalism was more likely. Rising productivity would allow a small group of workers to live in comfort as the great mass of global society suffered as superfluous humans, at best able to hope for occasional, precarious waged work. Without a powerful working class able to force firms to redistribute profits, Gorz was terrified of the use of automation to abolish work.

Leaving the decisions to AI corporations will make dystopia increasingly likely, especially as it continues to be used in policing and war, making some of Hollywood’s supposedly futuristic films seem frighteningly realistic. It is more common to think of AI as another more sophisticated form of technology that can adapt and change its actions to solve problems based on new information it can generate itself (see sidebar “Automation vs. AI vs. Regenerative AI”).

The Future of Work

While it is still far in the future, we have already seen glimpses of the dystopian future foretold in sci-fi films like “Blade Runner,” “Avengers: Age of Ultron,” “Ender’s Game,” and “RoboCop.” Perhaps no film tells this story better than “Elysium,” in which the rich barricade themselves on a space station protected by a wall of armed AI-robot soldiers with their boots on the necks of the destitute horde back on Earth. Humanity as unnecessary, destructive, and superfluous is a common theme in the genre. In “Avengers: Age of Ultron,” Ultron destroys humanity—and the superheroes—instead of saving it as it was designed to do. It is a fulfillment of scientist Stephen Hawking’s 2014 warning that “the development of full artificial intelligence could spell the end of the human race.”

Today, such visions of terror are the result of the lack of a coherent strategy to resist AI. In the European Union, the initiative has been seized by the social democratic parties and unions who are attempting to ameliorate and benefit from AI, rather than fight it. Think tanks, social democratic parties, and unions in the United States and European Union view AI as a question of privacy and safety, and have mostly approached it in terms of accountability, equity, inclusion, and transparency. This approach reduces a question of political economy to one of individual rights.

Automation vs. AI vs. Regenerative AI

In this article the terms “automation,” “AI,” and “generative AI” are used interchangeably. Automation can be any type of technology that replaces human labor by completing pre-programmed tasks with little to no human intervention. AI can be understood as a type of automation that uses available data to complete a task and can adapt its actions based on interpretations of new information and programming. Generative AI uses the existing data to create something new based on its capacity to make decisions, predict outcomes, and solve problems thus creating new data based on its analysis of patterns in the existing data used to program it. In short, AI interprets existing data but regenerative AI can use the data to create new data.

AI and the Rationalization of Labor

A class analysis of AI sees it as the latest type of technology designed to deskill and replace human labor. There are two simultaneous impacts of AI on both low-skilled and higher-skilled work. While automation uses programmed operations to allow a machine to complete a sequence of tasks, traditional AI uses data fed to it by programmers to make decisions or predictions based on its programming. A more advanced type of AI known as “generative AI” can use new information to autonomously adapt, change its actions, create something new, and even generate its own new data without the direct intervention of programmers. Both types of AI, like all forms of automation, are designed to reduce the dependence of capital on human labor.

In general, employers want to use AI to break our work into component tasks, a process referred to as the “rationalization” of labor, in order to determine which ones can be automated. Some types of complex tasks are already being accomplished by AI, leaving behind other tasks that cannot yet be automated to be done by human workers who augment the AI.

The debate is still ongoing about whether AI will augment or replace human labor. According to an International Labour Organization (ILO) study, the largest impact of generative AI will be in office work; many clerical and writing tasks will be automated, leaving behind only deskilled office workers to maintain machines running the AI, rather than doing the tasks the AI now accomplishes. The ILO study predicts that fewer high-wage and many low-wage jobs will be created to service the technology, with workers functioning as what Karl Marx and Friedrich Engels famously called “appendages of the machine.” While the ILO estimates a modest 48 million jobs will be automated worldwide, 181 million will be augmented by AI, with the greatest impact on work done by women.

Other estimates vary widely about the number of jobs that AI will allow to be fully or partially automated. In 2017, the corporate consulting firm McKinsey estimated that 60% of all jobs are vulnerable to some degree of automation, resulting in 400 to 800 million workers becoming obsolete by 2030. The Organization for Economic Cooperation and Development (OECD) reported that “27% of jobs are in occupations at high-risk of automation,” particularly in very skilled professions such as finance, medicine, and law.

Researchers from OpenAI, the same company that invented ChatGPT, warn that about 80% of U.S. workers could see as much as 10% of their work automated. In total, 19% of jobs have at least 50% of what they do at risk of automation. The OpenAI report established the dominant narrative among employers that “routine and repetitive tasks” are the most vulnerable to being automated by AI, but in fact, it is likely to impact more highly skilled types of work as well.

The result will be a higher rate of productivity by the workers still left whose work augments the AI. As a result, the consistent rise in productivity will continue to result in stagnant or declining wages as the productivity gains are passed along as higher profits. The ILO study estimates that “in the near future, generative AI systems similar to GPT are more likely to become productivity tools, supporting and speeding up the execution of some tasks within certain occupations.” A 2019 study by Daron Acemoglu of MIT and 11 other researchers found that AI was already making us work harder and produce more for less pay. As AI is introduced into the workplace, it reduces the number of human workers and human work hours, while increasing productivity for those workers who remain.

The issue is not whether AI will have an impact on work, but that it is already having one. A survey by the human resources and recruiting contractor Checkr found approximately 85% of workers have already used AI in their jobs. The question is how quickly that impact will intensify and spread. The already-widespread use of algorithms, learning-management systems, and automated software functions such as spell checks, internet searches, and chatbots in our work is an indicator of how rapidly we can expect more AI in the workplace.

Traditional AI has already had a large impact on manual labor, including many kinds of service work and manufacturing, for several decades—a trend that will no doubt continue. Now, generative AI has begun to replace more complex types of “cognitive” work. KPMG’s U.K. labor market study (see table) projects that authors, writers, and translators will have 43% of their tasks automated in the next decade. Programmers, software developers, PR and communications workers, and IT support are estimated to have 20% or more of their work tasks affected. Scientists and higher education faculty were the lowest on the list, but still face 6% of their work being automated by AI.

A Future Dystopia

Whether AI augments, deskills, or makes some kinds of human workers obsolete, it will likely bring the Global North into much closer alignment with the rest of the world, where formal employment is precarious, tedious, and poorly paid. Working for a specific boss, in a single workplace, and with regular pay has long been nonexistent for most of the human population. According to the ILO, in Africa, 85.8% of people have no formal jobs. That number is 68.2% in Asia and the Pacific, 40% in the Americas, and 25.1% in Europe and Central Asia. In the United States, estimates for precarious labor range widely. The U.S. Bureau of Labor Statistics says it comprises 10.1% of the workforce. McKinsey estimates it at 32%.

These developments are hardly surprising. In 1867, Marx already imagined the automated factory run by self-learning AI. In Capital Volume I, he wrote that “As soon as a machine executes, without man’s help, all the movements required to elaborate the raw material, and needs only supplementary assistance from the worker, we have an automatic system of machinery, capable of constant improvement in its details.”

The focus on AI obscures not only the loss of social control when human labor is rendered obsolete, but why the technology is used in the first place. Just like the National Eating Disorders Association’s misstep with AI or the Hollywood studios’ desire to use AI writers and actors, bosses everywhere attempt to reduce their reliance on workers by augmenting or replacing them with new productivity-increasing technology. They are especially keen to do so when workers attempt to organize and assert their power.

Bosses often introduce technology in response to worker organizing, disruptions, and strikes. As Marx explained in Capital Volume I, “It would be possible to write a whole history of the inventions made since 1830 for the sole purpose of providing capital with weapons against working-class revolt.” AI is just the latest technology to play this role.

Because high unemployment risks social disorder, capitalists must keep people working to some extent even as they try to reduce their reliance on human labor in order to increase profits and maintain control of their firms. If AI can do the work of many human workers, it threatens to blow up the use of work as social control. The introduction of labor-saving technology to reduce the reliance on human workers while making the fewer workers who keep their jobs work harder is known as the “zerowork paradox.” AI replaces workers, and those who still have jobs work harder until they, too, can be automated. AI will gradually automate parts of work so that capital will need fewer and fewer workers. The robot apocalypse is not that robots will take our jobs, it is that work will no longer keep us under control.

Getting rid of human workers without a new means of imposing social control means not only disorder but also brutal totalitarian repression. For Gorz, that meant “what is being preserved is not the capitalist system but capitalism’s system of domination.”

This is what makes the totalitarian police states of sci-fi films like “Elysium” and “RoboCop” more realistic, and thus more terrifying. In the bleak dystopian future, the human population itself becomes entirely unneeded.

A New Social Contract

A more hopeful AI future would promise less work, a better quality of life, and even the abolition of work itself. As we produce more in less time and with less effort, some people believe that the increased fruits of our collective labor could be shared globally to meet all human needs. Visions of what that would look like have been the focus of recent books such as Peter Frase’s 2016 book Four Futures: Life After Capitalism and Aaron Bastani’s 2020 book Fully Automated Luxury Communism: A Manifesto, advocating for humans to exploit the work of robots to fund a life of human leisure. Whichever outcome occurs won’t come to pass without a struggle for power to either control the fruits of the capitalist economy or rupture it to get beyond capitalism.

This May, I interviewed Boston College professor Juliet Schor, author of the 1992 book The Overworked American: The Unexpected Decline of Leisure. She said new technology in the workplace has not always led to higher unemployment. Now is not the first time we have faced mass unemployment and destitution because of the potential replacement of human labor with technology. The steam-powered factory, the railroads, and automobiles made many types of agricultural and skilled work obsolete, threw many people into joblessness and precarity, and coincided with a series of depressions with high unemployment between the 1860s and 1930s. Workers responded by organizing unions, striking, and even arming themselves in self-defense against corporate thugs, judicial injunctions, the army, state militias, the National Guard, and police—as I show in my first book, When Workers Shot Back, which was published in 2019.

Schor proposes that we advocate for converting rising productivity from AI into shorter hours, and to get rid of the worst kinds of work: “The key is that workers are organized. Get the shorter hours ball rolling now, for the movement for a four-day week. People want this a lot now.”

She points to four-day work week trials now under way in the United Kingdom and Ireland, which she has been involved in, and at Samsung in South Korea. The problem, according to Schor, is that we are not democratically making the necessary choices about whether and how we want to grow the economy. Nor are we confronting the ecological and social damage from doing so. This is the paradox of capitalism. We work to survive even though work is a primary source of the threat of ecocide.

Schor’s sense of the possibilities is rooted in the knowledge of the outcomes of previous worker struggles. Capitalists responded by introducing new technology to deskill, replace, or discipline insurgent workers by making them work harder and produce more. But it also led to two other different, conflicting outcomes. These struggles raised union density, led to disruptive strike waves, and extracted social democratic reforms. In countries with strong unions and labor parties, these advances were enshrined in both the shortening of the work week and earlier retirements. In countries like the United States, it resulted in the 1938 Fair Labor Standards Act that provided the 40-hour work week for some workers, and in labor union contracts that traded higher wages for even more intensive work.

The results of this deal can be seen in historical declines in work hours. According to Our World in Data (see figure), between 1870 to 2017, annual work hours declined about 40% worldwide, from about 3,100 to about 1,800. Work hours declined by more than half in five European countries, the United States, and Australia. However, downward progress stalled in nearly every country since the 1970s. Since the demand for the legal eight-hour workday was enshrined into law for most hourly workers in the Fair Labor Standards Act, our unions have largely ignored the issue of work, who controls it, and how much of it we do.

After the 1950 “Treaty of Detroit” between the United Auto Workers and General Motors, it became common for unions to cut deals with the boss to produce more if workers get paid a larger share of the profits our labor produces. That closed the inequality gap for some white male workers, but the effects didn’t last. The boss responded with systemic racism and sexism, privatization, outsourcing, bankruptcy, mergers, and automation—and now AI.

A Democratically Controlled Economy

The future of AI will be the outcome of the balance of power between workers and capitalists. Rising working-class power that threatens to disrupt the economy is the most likely way to force the question of whether we ban AI or regulate who controls the technology and what it produces. This is the reason why the screenwriters and actors went on strike over AI.

The WGA strike resulted in a new contract that empowers the workers to control AI. They now get to decide how AI is used in writing scripts. The contract requires the companies to disclose which materials are produced by AI and a screenwriter cannot be forced to use those materials. When writer’s materials are used to “train” AI, the union can challenge it as a contractual violation. Most media handwringing about AI and the future of work fails to address the fact that it is not only a question about work but more fundamentally about how we organize society. The future will turn on how we decide to govern ourselves and what kind of economic system we want—and organized workers will have the power to decide it. This means both asking hard questions and taking action, both of which we are not yet doing.

Will we implement a new social contract that redistributes income and wealth in exchange for fewer hours of work at safer and more fulfilling jobs? Will we rise up and abolish the life-killing capitalist system and democratically seize control of and run the economic system in order to abolish work as a coerced activity for survival?

Ultimately, our survival will be determined by our ability to democratically control work and the economy. Too many of us work far too much for far too little. Meanwhile, too many have too little or no work. This is the irrationality of work under capitalism: it really does not determine whether we survive or not.

Getting past capitalism would allow us to democratically decide what work is needed and what is not. That process will involve sorting out what work is destructive and what is needed to take care of one another and meet human needs. Abolishing work as a coercive means of control, exploitation, and domination would free the full range of human creativity and expression. This is the only way to avoid both the robot apocalypse and ecocide. The answer to the question about the future of work is that work has no future.

Answering the question about the future of work, as well as AI, means answering the question of the future of humanity and the planet. Failure to answer either question will make the AI work apocalypse more certain. We still have a little bit of time to determine whether the metaphorical robots become the new slave drivers of the elites, used to keep the unneeded humans in line. The actors and screenwriters who struck over control of AI have reminded us that our most important weapon against AI is already in our repertoire. We need to use it well.

teaches at UC Berkeley and San José State University. He is editor of Workers’ Inquiry and Global Class Struggle (Pluto), and the author of When Workers Shot Back (Haymarket) and the new book We the Elites: Why the U.S. Constitution Serves the Few (Pluto). Follow him at @OvetzRobert

Dominic Patten, “SAG-AFTRA Strike Could Hinge On AI; Deep Divisions Remain Between Actors & Studios In Final Hours Of Talks,” Deadline, July 10, 2023 (; Mike Elk, “160,000 SAG-AFTRA Members to Strike with 11,000 of the Writers Guild for 1st Time in 60 Years,” Payday Report, July 13, 2023 (; Chloe Xiang, “Eating Disorder Helpline Fires Staff, Transitions to Chatbot After Unionization,” Vice, May 25, 2023 (; “NEDA Suspends AI Chatbot for Giving Harmful Eating Disorder Advice,”, June 5, 2023 (; Gabriel A. Silva, “The ChatGPT Debate: Are We Intelligent Enough To Understand ‘Intelligence’?,” Forbes, March 14, 2023 (; Dominic Patten, “SAG-AFTRA Strike Could Hinge On AI; Deep Divisions Remain Between Actors & Studios In Final Hours Of Talks,” Deadline, July 10, 2023 (; ChatGPT (; Cody Godwin, “Burger-Flipping Robot Begins First Shift,” BBC, March 5, 2018 (; Zeynep Tufekci, “Failing the Third Machine Age: When Robots Come for Grandma,” Medium, July 22, 2014 (; Robert Ovetz, “Taylor’s Digital Stopwatch: What the U.S. Labor Movement Can Learn from European Workers Who are Organizing Against ‘Algorithmic Management,’” Dollars & Sense, September/October, 2022 (; Nitin Kumar, “Legal Tech: Artificial Intelligence-Enabled Review,” Forbes, August 23, 2021 (; Jacob Zinkula, “AI is Helping Your Company Decide Who to Lay Off,” Business Insider, February 23, 2023 (; André Gorz, Paths to Paradise: On the Liberation from Work, London: Pluto Press, 1985; Center for AI Safety, “Statement on AI Risk,” n.d. (; “Stephen Hawking Warns Artificial Intelligence Could End Mankind,” BBC, December 2, 2014 (; Gerard Rinse Oosterwijk, “AI, platforms and (human) workers’ rights,” Social Europe, July 7, 2023 (; OECD Employment Outlook: Artificial Intelligence and the Labour Market, 2023 (; Sam Altman, “Planning for AGI and Beyond,” OpenAI, February 24, 2023 (; Cecilia Kang, “OpenAI’s Sam Altman Urges A.I. Regulation in Senate Hearing,” New York Times, May 16, 2023 (; Karl Marx and Friedrich Engels, Manifesto of the Communist Party, London: Pluto Press, 1848[2017] (; James Manyika, Susan Lund, Michael Chui, Jacques Bughin, Jonathan Woetzel, Parul Batra, Ryan Ko, and Saurabh Sanghvi, “Jobs Lost, Jobs Gained: What the Future of Work Will Mean for Jobs, Skills, and Wages,” McKinsey Global Institute, November 28, 2017 (; OECD Employment Outlook: Artificial Intelligence and the Labour Market, 2023 (; Dan Milmo, “AI Revolution Puts Skilled Jobs at Highest Risk, OECD Says,” The Guardian, July 11, 2023 (; Tyna Eloundou, Sam Manning, Pamela Mishkin and Daniel Rock, “GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models,” arXiv:2303.10130, Cornell University (; Yael Selfin and Dennis Tatarkov,“Generative AI and the UK Labour Market,” KPMG, June 2023 (; PaweĊ‚ Gmyrek, Janine Berg, David Bescond, “Generative AI and Jobs: A global analysis of potential effects on job quantity and quality,” ILO 2003 (; Daron Acemoglu, Gary W. Anderson, David N. Beede, Cathy Buffington, Eric E. Childress, Emin Dinlersoz, Lucia S. Foster, Nathan Goldschlag, John C. Haltiwanger, Zachary Kroff, Pascual Restrepo and Nikolas Zolas, “Automation and the Workforce: A Firm-Level View from the 2019 Annual Business Survey,” National Bureau of Economic Research Working Paper Series (; Sarah Kessler, “The A.I. Revolution Will Change Work. Nobody Agrees How.,” New York Times, June 10, 2023 (; Sara Korolevich, “Insights from American Workers: A Comprehensive Survey on AI in the Workplace,”, May 24, 2023 (; Yael Selfin and Dennis Tatarkov,“Generative AI and the UK Labour Market,” KPMG, June 2023 (; Daron Acemoglu, David Autor, Jonathon Hazell and Pascual Restrepo, “AI and Jobs: Evidence from Online Vacancies,” National Bureau of Economic Research Working Paper Series (; ILO News, “More than 60 Per Cent of the World’s Employed Population are in the Informal Economy,” April 30, 2018 (; Steven F. Hipple and Laurel A. Hammond, “Self-employment in the United States,” U.S. Bureau of Labor Statistics, March 2016 (; André Dua, Kweilin Ellingrud, Bryan Hancock, Ryan Luby, Anu Madgavkar, and Sarah Pemberton, “Freelance, Side Hustles, and Gigs: Many More Americans have Become Independent Workers,” McKinsey & Company, August 23, 2022 (; Karl Marx, Capital, Vol. I, NY: Vintage, 1977; Zerowork (; Sam Altman, “Planning for AGI and Beyond,” OpenAI, February 24, 2023 (; “Elysium,” 2013 (; Hannah Arendt, The Origins of Totalitarianism, NY: HBJ, 1951, Robert Ovetz, When Workers Shot Back, Chicago: Haymarket Press, 2019 (; Charlie Giattino, Esteban Ortiz-Ospina and Max Roser, “Working Hours,” Our World in Data, December 2020 (; John Maynard Keynes, “Economic Possibilities for our Grandchildren,” in Essays in Persuasion, New York: W. W. Norton & Co., 1963; Justin, “UAW History—The Treaty of Detroit,” July 25, 2020 (; Thomas Paine, Agrarian Justice, 1797 (; Kevin Van Meter, “We Should Demand Democratic Workplaces, But What Does That Mean?,” New Politics, December 26, 2022 (; Robert Ovetz, “Hollywood Screenwriters Strike Put a Leash on AI,” The Chief, October 17, 2023 (

Did you find this article useful? Please consider supporting our work by donating or subscribing.

end of article