AI has entered the workplace with all the subtlety of a wrecking ball. So far, most of the headlines pound home what workers stand to lose: jobs, privacy, and the last shreds of autonomy the modern workplace still affords. From hiring tools that filter out candidates using biased data to productivity trackers that log every keystroke, the technology has largely been cast as the boss’s new weapon.
And for good reason. But the assumption that AI will inevitably serve the interests of the companies building it deserves more scrutiny than it gets. In recent years, generative tools like ChatGPT and DALL·E have made the technology’s power visible to millions, while leaps in computing power and data have made these systems far more capable. Against the backdrop of a post-pandemic economy marked by rolling layoffs and deepening inequality, AI has shifted from a background efficiency tool to a frontline driver of economic and political change. A small class of executives and investors have set the terms of AI’s development, and they would prefer you to mistake that arrangement for inevitability.
In the hands of those profiting from labor cuts, AI becomes a devastating instrument of exploitation. Built to serve shareholders, it behaves like any other profit-maximizing tool in history.
But this transformative technology now used to surveil and deskill workers could, in theory, be used to do the opposite: strengthen protections, expedite public services, and expand access to life-saving resources. As youth activist Sneha Revanur contends, “If these algorithms were programmed for good, they could be used for good.”
AI is not inherently exploitative and shouldn’t be treated as such. It mirrors the values and intentions of its creators. So when those creators are a homogeneous group of venture-backed technologists insulated from the consequences of their own products, AI reproduces their blind spots at scale. The core question remains: who decides what AI is for?
AI as a Public Utility
An open-source AI run by a city agency, workers union, or nonprofit, governed transparently and trained on public data, would answer to an entirely different set of priorities than the profit-driven systems currently dominating the market.
In Barcelona, residents already use Decidim to propose, debate, and vote on policies, while Project DECODE gives them ownership and control of their personal data. Both sit within a wider network of smart-city systems that optimize traffic flow, coordinate bus networks, and enhance emergency response in real time. Crucially, Barcelona built all of these tools under public governance, not as a concession to tech companies but as an assertion of democratic authority over digital infrastructure. It is the kind of reckoning that musician Björk has argued every new technology demands. “You always have to figure out the morality [of a new technology], and what it means on every level: socially, personally, and politically.”
Applied to workforce issues, this model could match workers to apprenticeships based on skills and interests rather than connections, route surplus food from restaurants to shelters through real-time coordination, or analyze workplace injury data to flag dangerous patterns before accidents happen. Within this framework, AI stops functioning as a tool for extraction and becomes infrastructure for equity, designed to serve the people who depend on it most.
The Unionized Algorithm
If corporations can train AI to monitor workers, unions can train it to defend them. The same warehouse algorithms that track productivity quotas could just as easily flag unsafe workloads, scheduling systems built to maximize coverage could catch violations of fair scheduling laws, and the performance-review data that companies already collect could be audited for systemic bias in promotions.
On the ground, worker-led tech initiatives are translating this idea into concrete results. Through Coworker.org‘s digital tools, Starbucks and REI employees have won tangible victories in organizing campaigns. Across Europe, IG Metall is negotiating transparency and safeguards in algorithmic management, while in the U.S., the Athena Coalition has backed the federal Warehouse Worker Protection Act to bring quota transparency, limit surveillance, and improve safety at Amazon and similar employers.
For the Berkeley Labor Center, the answer lies in legislative reform. “Technology is not inherently good or bad, but neither is it neutral; public policy must ensure technology serves and responds to workers’ interests.” If workers secure the tools and leverage to determine how AI is governed in the workplace, the technology shifts from an invasive control mechanism to a foundation for better conditions on the job.
AI That Cuts Red Tape Instead of Jobs
Across the U.S., public benefits systems are buckling under chronic understaffing and outdated technology, forcing applicants for unemployment insurance, disability benefits, housing assistance, and immigration aid to wait months for help that should arrive in days. The devastating consequences, from missed rent payments to lapsed medical coverage, fall hardest on the people these systems were built to serve.
Under proper public oversight, AI could help break these bottlenecks by doing what decades of budget cuts and hiring freezes have prevented. In New Jersey, an AI translation assistant developed with Google.org has tripled translation speed for unemployment insurance forms, improving access for Spanish-speaking claimants and making its tools available to other states. In Nevada, an AI prescreening tool is reducing backlogs and processing claims with 99.99% accuracy. Beyond individual programs, AI can translate documents into dozens of languages in seconds, verify paperwork, and guide applicants through complex processes, escalating only the most difficult cases to human staff.
AI can translate documents into dozens of languages in seconds, verify documents, and guide applicants through complex paperwork, escalating only complex cases to human staff.
Used well, this technology could resolve claims in days rather than months, match housing applicants to available units in real time, and streamline aid delivery while safeguarding privacy and upholding independent oversight. But government AI deployments have also produced catastrophic failures, from automated benefits systems that wrongly denied thousands of legitimate claims to predictive tools that replicated the very biases they were supposed to eliminate. As New York Times columnist Julia Angwin observed, AI is “looking less like an all-powerful being and more like a bad intern whose work is so unreliable that it’s often easier to do the task yourself.” The technology’s potential in public services is real, but only if it is built with the kind of rigorous oversight and democratic accountability that the private sector has shown little interest in providing.
AI as Corporate Watchdog
If the same analytical power currently trained on workers were redirected at those with real decision-making authority, corporations would find themselves under the kind of scrutiny they have long reserved for their employees. AI could analyze hiring and promotion patterns to detect bias, monitor pay data for hidden wage gaps, flag environmental violations, and map the lobbying networks that shape policy behind closed doors. ProPublica has already used such tools to reveal how risk-assessment algorithms in criminal sentencing rated Black defendants as higher risk despite similar profiles, exposing deep algorithmic bias. Global Witness has applied AI to identify fossil fuel lobbyists at COP climate talks, revealing whose interests were shaping climate negotiations.
Instead of functioning as the boss’s microscope, AI could become the public’s telescope for institutional behavior, magnifying consequential patterns and keeping them in plain sight. It would not replace journalists, regulators, or advocacy groups, but it could equip them with faster, deeper insights to act on. The shift from tracking worker “efficiency” to tracking corporate responsibility would be straightforward in principle but seismic in impact.
The Mutual Aid Machine
When traditional systems buckle, civic technologists and volunteer developers have a decent track record of producing solutions that are scrappier, more targeted, and more responsive to the communities they serve. During the 2020 pandemic, a crowd-sourced mask map in Taiwan evolved from a simple availability tracker into a real-time, pharmacy-linked distribution system that became a model for public health logistics. In the aftermath of earthquakes, volunteer networks have used tools like AIDR and MicroMappers to rapidly verify and map critical needs across affected regions, often outpacing the official response.
That same grassroots impulse is rewriting tenant justice, an area where the gap between legal rights and the ability to exercise them has long been one of the most glaring inequities in American housing. At Cornell Law, students launched Teny, a chatbot that delivers legal guidance on landlord disputes to renters in upstate New York. In New York City, the AI-powered Roxanne the Repair Bot, developed by Housing Court Answers and NYU, uses conversational logic to help tenants document repairs, draft letters, and file legal reports, turning what used to require a lawyer’s time into something a renter can do from a phone.
Each of these projects was built from the ground up around a specific community need, instead of a revenue model or a pitch deck. They match aid faster than bureaucracy, translate legal rights into accessible tools, and demonstrate that the technology for a more equitable distribution of resources already exists. The missing ingredient has never been capability, but rather the commitment to building for the people who stand to benefit most.
The Clock Is Ticking
The window for shaping AI governance is still open, but it is closing faster than most people realize. As computer scientist Timnit Gebru warns, “Right now, only a handful of people and organizations have the power and resources to automate decision-making.” With every system built to monitor workers or eliminate jobs, with every dataset collected without consent and every algorithm deployed without transparency, the space for democratic intervention narrows further.
Throughout this piece, evidence has shown that AI designed to serve communities rather than shareholders produces better outcomes for everyone. Barcelona’s citizen-led platforms, union-backed watchdog systems, tenant justice chatbots, and grassroots crisis response networks all deliver proof of concept. But without sustained, organized pressure from workers, communities, and researchers willing to challenge the concentration of AI power, these models will remain exceptions rather than the standard.
Designed for the many, AI can be a force for shared prosperity. Captured by the few, it will function as an enforcer for the 1%, deepening the very gaps it could potentially rectify. The time to decide which version prevails is now, before the cement sets.
