AI Has a Labor Problem

The World Bank has high hopes for the positive impact of AI. Its flagship World Development Report this year is even titled “The Promise of Artificial Intelligence.” 

From its annual meetings in Bangkok through Oct. 18 to the Global AI and Digital Summit in Seoul next week, the bank is promoting AI’s potential to deliver “more and better jobs.” But what would it take to make that promise real for workers? 

The experience of platform workers offers a cautionary tale. For millions of gig workers worldwide, algorithms, including AI systems, already act as the boss. 

Human Rights Watch research in the United States found that major companies use opaque and frequently changing algorithms to set pay, leaving workers unable to understand how their earnings are calculated. Customer ratings feed into systems that determine access to jobs, with little opportunity to contest unfair scores. Workers whose accounts are deactivated struggle to get clear reasons or reach anyone with the authority to reverse the decision.

The problem goes beyond opaque decisions and limited ways to challenge them. Companies also insist that workers are mere independent contractors to avoid long-established employer responsibilities, despite the control they exercise over workers.

Take Uber, which now operates in over 70 countries. Uber generally treats drivers as self-employed, saying that it merely connects them with customers. Yet its systems set fares and govern access to work. Treating drivers as independent contractors rather than employees leaves them without minimum wage guarantees or employer contributions to social insurance. The company directs the work while shifting much of its cost and risk onto workers.

Our research across platforms and countries, including in Lebanon, Mexico, Pakistan, the U.K., Kenya and India, as well as across the Persian Gulf, found striking similarities in platform workers’ experiences. They showed us pay records revealing sharp daily fluctuations in earnings and detailed substantial costs for vehicles, fuel and other work equipment. On some days, their expenses exceeded their earnings. Some workers injured on the job spent months without income, relying on family and friends or taking on debt to survive.

A delivery worker in Delhi told us he faced financial penalties for missing delivery targets. While rushing to complete an order, he braked to avoid a collision with a motorcyclist, fell and broke his arm. He paid for his medical treatment himself. “We don’t get anything from the company,” he said. 

His experience shows how this system compounds the harm to workers. Algorithmic management pressures them to take risks and punishes them if they fail to comply, while they bear the consequences if the risks cause them to get hurt. 

This rapidly expanding and precarious system requires robust regulation of how companies use technology to manage work and accountability for workers’ pay, safety and social security. Such regulation should, at a minimum, ban algorithmic management practices that pose an unacceptable risk to workers’ rights.

A new international labor convention offers hope that these protections can become a reality. In June, governments, workers and employers at the United Nations’ International Labour Conference adopted Convention No. 193 on Decent Work in the Platform Economy by a vote of 406 to 8. It is the first binding international labor standard to regulate algorithmic management for platform workers.

The convention requires ratifying governments to ensure that workers can request written explanations of significant automated decisions that adversely affect their work. Workers must have access to review, with appropriate human involvement, when such decisions withhold their pay or cut their access to work. It also provides safeguards for pay, safety, social security and the right to organize, although some protections depend on employment status. 

Crucially, governments must take measures to ensure correct employment classification, guided mainly by how work is actually performed and paid rather than by the labels a company chooses to apply in a contract.  

Together, these provisions address both how technology is used to manage workers and companies’ responsibilities toward them. They give the World Bank a concrete yardstick for some of the “better jobs” it hopes AI will deliver.

The stakes extend beyond ride-hailing and delivery. Research by the group WIEGO (Women in Informal Employment: Globalizing and Organizing) highlights that that gig-ification has spread into domestic work, beauty services and data work, with many of these workers also facing algorithmic control and limited ways to resolve workplace problems. The International Labour Organization and European Commission have also documented algorithmic management in conventional workplaces, including logistics and healthcare. Workers across these sectors should also be entitled to protection from harmful uses of these systems and effective ways to challenge decisions.

Governments should ratify the new convention, implement it with workers’ participation, and address harmful uses of automated management across labor markets. The World Bank should support these efforts through its advice and financing, helping governments to enforce labor protections and to ensure that workers can challenge decisions affecting their pay or access to work.

In Bangkok and Seoul, “better jobs” should mean a living wage, protection when illness or injury stops someone from working and the power to negotiate working conditions. These are rights that require enforcement. The institutions promoting AI have a responsibility to help make that promise real.