How will artificial intelligence (AI) impact the workplace? To answer that, we must have a much clearer sense of what AI is. In the most basic sense, AI is software — powerful software. It uses a mix of code and layers of other software programs that have already run trillions of calculations on existing text and images to generate predictions. AI uses these calculated responses to spit out pretty good answers to almost any prompt asked of it. Think of it as a cross between a web search and the person at work who has an answer for anything — whether they know the right answer or not. It is built to mansplain with certainty.

Why does this less-flattering definition of AI matter to how it will impact work? Workplaces are more than the sum of their parts. Sometimes there are obvious answers to questions at work. Most of the time, workers figure out the most creative path to get something done. Whether delivering a couch, taking a patient’s pulse, or building a car, workers are constantly collaborating and making snap judgments — predictions — about the best action to take next. The most productive workplaces do this by setting up systems so that workers can draw on their collective experience, capabilities, and diverse points of view. Individual workers figure out the best way to get something done, together, often without upper-level management even knowing.

Studying workplace productivity and recognizing the social dynamics of individual decisions is not new. Frederick Winslow Taylor, the father of scientific management, focused on improving operational efficiency by using performance analysis to inform iterative workflow improvements. He broke down each manual job into its individual physical motions, analyzed these to determine which were essential, and timed the workers with a stopwatch to produce data-driven analyses about which movements seemed like a “waste of time.” Eliminating unnecessary motion, the worker follows a machinelike routine and becomes far more productive — in theory. In practice, Taylor’s recommendations could not account for real-world problems (e.g., broken drill presses) and practical workarounds (e.g., workers swapping stations) that no preordained routine could accommodate.  

Unlike Taylor’s close studies that analyzed workers’ discrete, isolatable tasks, AI in the workplace will need masses of data to replicate the analytics of Taylor’s deep, micro-level studies. Here’s the open secret of AI: it is nothing without two key ingredients — gob-smacking amounts of data and a lot of computing power to draw on models of prior decision-making and offer reliable responses to users’ requests for a prediction.

Large language models (LLMs) — e.g., OpenAI’s ChatGPT, Microsoft’s Copilot for Office, and Google’s Gemini for Google Workplace — are powerful tools for workplace productivity. But they are more than that. They are derived from myriad examples of other people’s prior decisions: What word to type next? What image should be put on the left or right of a slide?     

While others have generally called for more worker voice in the development of AI, it will also be essential that workers have the power to control and bargain over workplace productivity data in particular. Workers’ voices are important, yet most AI tools, drafting off LLMs, are built from repeatedly scraping the internet and using other data sources with no direct connection to a specific worksite, let alone a work team’s input. The driving force behind Taylorism — the need for companies and bosses to collect more individual output and find efficiencies — also drives the development of AI. Most digital systems produce workplace data that good bosses can use to reward individual workers, and bad bosses can use to monitor and fire workers under the sign of “productivity.”

A group of community health worker organizations in California is currently showing us what it looks like to center workers’ voices. They are working with my research team of social scientists and technologists to create blueprints for new ways to store and learn from community members’ sensitive data that bring the full force of privacy techniques and data analytics to their care work. They are invested in stewarding community-operated data trusts that can protect and govern local data in transparent and accountable ways. Rather than handing data over to outside institutions, these small organizations on the front lines of public health are exploring how to turn their shared records of local support and information exchange into key insights for how to best manage the next pandemic or climate disaster. They are prioritizing revocable consent that gives clients the power to audit what happens to their data and privacy-preserving approaches that make community-run data trusts functional health and human service cooperatives rather than sites of data extraction.        

We can look to the European Union for work councils as a model for bringing workers’ needs into the mix. Microsoft and other US-based multinational companies routinely collaborate with workers councils in the EU that have expectations around protecting the rights of customers — and workers — as tools are built for them. But there is no reason to wait until AI productivity tools are near release to bring workers’ expertise to the table. What would it look like if not only workers’ voices were heard, but their lived experiences and expertise were baked into the AI itself?     

Workers need control, the right to collectively bargain, and the right not to have their social interactions reduced to an output for a boss. This is the biggest difference between Taylorism and now. Taylor timed workers’ every move, trying to shave seconds off the clock ticking down to assemble a car part. While he aimed to incrementally improve the aggregate productivity of individual workers, today’s AI tools are not directly or closely looking at individual output; they collect and statistically slice and dice every interaction we have at work before deciding which type of work (or worker) might be on the chopping block. No workplace should annihilate our social selves in the name of productivity.

Policies are needed to guarantee that workplace activity data is built for collective bargaining that prioritizes what workers need, not what bosses want. We need to keep work from becoming a surveillance site in the name of efficiency. Research actually shows that productivity decreases as employee monitoring increases, especially in creative and teamwork environments. We also need to make it illegal for employers to track individuals outside of the workplace. 

Workplace productivity data is the most valuable bargaining chip workers could grab hold of, especially in the moment when regulation and labor protections seem more fragile than ever. AI will either become hard-to-counter workplace surveillance in the name of productivity for the employer or workplace activity for worker power.

AI models built on workers’ productivity data will make better tools as they will reflect workers’ shared rather than individual labor. Work productivity tools that emphasize workers’ participation in co-designing the priorities of AI applications and what data are used to train these models should be the new paradigm for how workers make choices about where technology goes for the future of work.

About the Author

Senior Principal Researcher,
Microsoft Research

 

Gray is Senior Principal Researcher at Microsoft Research and Faculty Associate at Harvard University’s Berkman Klein Center for Internet and Society. She maintains a faculty position in the Luddy School of Informatics, Computing, and Engineering with affiliations in Anthropology and Gender Studies at Indiana University. Mary, an anthropologist and media scholar by training, focuses on how people’s everyday uses of technologies transform labor, identity, and human rights. Mary earned her PhD in Communication from the University of California at San Diego in 2004, under the direction of Susan Leigh Star. In 2020, Mary was named a MacArthur Fellow for her contributions to anthropology and the study of technology, digital economies, and society.

Mary’s work includes In Your Face: Stories from the Lives of Queer Youth (1999) and Out in the Country: Youth, Media, and Queer Visibility in Rural America (2009), which looked at how young people in rural Southeast Appalachia use media to negotiate identity, local belonging, and connections to broader, imagined queer communities. The book won the American Anthropological Association’s Ruth Benedict Prize and the American Sociological Association’s Sexualities Studies Book Award in 2009. And, with Colin Johnson and Brian Gilley, Mary co-edited Queering the Countryside: New Directions in Rural Queer Studies (2016), a 2016 Choice Academic Title.

In 2019, Mary co-authored (with computer scientist Siddharth Suri), Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass. The book chronicles workers’ experiences of on-demand information service jobs—from content moderation and data-labeling to telehealth—work that is essential to the global growth of artificial intelligence and platform economies more broadly. It was named a Financial Times’ Critic’s Pick and awarded the McGannon Center for Communication Research Book Prize in 2019. The book was also awarded the 2020 Communication, Information Technologies, and Media Sociology section of the American Sociological Association (CITAMS) Book Award Honorable Mention. The book has been translated into Korean and Chinese.

Mary chairs the Microsoft Research Ethics Review Program—the only federally-registered institutional review board of its kind in Tech. She is recognized as a leading expert in the emerging field of AI and ethics, particularly research at the intersections of computer and social sciences. She sits on the editorial boards of Cultural Anthropology, Television and New Media, the International Journal of Communication, and Social Media + Society. Mary’s research has been covered by popular press venues, including The Guardian, El Pais, The New York Times, The Los Angeles Times, Nature, The Economist, Harvard Business Review, The Chronicle of Higher Education, and Forbes Magazine. She served on the Executive Board of the American Anthropological Association and was the Association’s Section Assembly Convenor from 2006-2010 as well as the co-chair of the Association’s 113th Annual Meeting. Mary currently sits on several boards, including the California Governor’s Council of Economic Advisors, Public Responsibility in Medicine and Research (PRIM&R), and Stanford University’s One-Hundred-Year Study on Artificial Intelligence (AI100) Standing Committee, commissioned to reflect on the future of AI and recommend directions for its policy implications.

---

About this Series

This is part of a series called “Back to the ‘Future of Work’: Revisiting the Past and Shaping the Future,” curated by the Aspen Institute Future of Work Initiative. For this series, we gather insights from labor, business, academia, philanthropy, and think tanks to take stock of the past decade and attempt to divine what the next one has in store. As the future is yet unwritten, let’s figure out what it takes to build a better future of work.

About the Future of Work Initiative

The Future of Work Initiative seeks to build and disseminate knowledge rooted in workers’ experiences. We aim to advance policy ideas at the local, state, and federal level, backed by evidence. And we strive to build community and activate leaders to carry these conversations forward across sectors and around the globe. This initiative was founded in 2015. Recognizing the need for a comprehensive approach to building a more inclusive economy, the Future of Work Initiative integrated with the Aspen Institute Economic Opportunities Program in 2021.