During the last big wave of automation in the 1980s and 1990s, technology produced new jobs and made others obsolete. The demand for rote-labor workers had diminished, while that for workers with computer-based skills had gone up. Laborers who didn’t have much experience beyond their rote jobs were, in turn, hit the hardest, and those laborers tended to be black: “Even before the economic restructuring of the nation’s economy,” wrote William Julius Wilson in his 1996 book When Work Disappears, “low-skilled African-Americans were at the end of the employment queue.” Who will the biggest victims be in this new age of automation, in which artificial intelligence dominates and tasks such as driving are computerized?

Automation isn’t inherently a boon for those with means. As Byron Auguste—the CEO and co-founder of Opportunity@Work, an organization that connects people who lack credentials to jobs—said on a Brookings Institution panel earlier this month, it doesn’t “naturally empower the top of the economic pyramid.” Nor, he argued, is automation like gravity—a force of nature that can’t be stopped. Humans have control over how technology is spread and implemented across industries—and they’re driven by what Auguste described as “economic incentives” and “cultural assumptions.” These incentives and assumptions indirectly determine automation’s impact on the workforce. And in this new wave, Latinos are the main casualties.

Americans favor assigning to robots jobs that are dangerous and unhealthy for humans rather than those that require human sensibilities (such as caregiving and driving), according to a recent study by the Pew Research Center. Latinos, especially Latino men, are heavily overrepresented in those challenging, and oftentimes repetitive, roles.

The 20 most popular occupations among Latinos in the U.S. virtually all involve hard, manual work and are often dangerous, according to the Brookings Institution fellows Mark Muro and Jacob Whiton, citing jobs in agriculture, roofing, and construction specifically. Construction, for example, had the most fatal work injuries among the Bureau of Labor statistics categories. Latinos make up a whopping 63 percent of drywall installers, a dangerous job because of the harmful irritants in drywall dust, while 30 percent are white and 7 percent are African American, according to Muro and Whiton’s calculations. Latinos are also overrepresented in repetitive work that requires few digital skills—such as that in hospitality—and, according to research by the Mckinsey Global Institute, is most susceptible to automation using currently available technologies.

Automation in the aforementioned sectors has already begun. That’s in part because many of them have lots of job openings (the construction sector, for example, has nearly 200,000 vacancies) that employers could eliminate using robots and other technology, whose efficiency and productivity would also save the companies money in the long-term. Drones, for instance, can take inventory on a site faster than humans can. But the automation of jobs such as drywall installation and roofing is also appealing because the positions are so dangerous: Replacing humans with robots in those jobs could prevent countless injuries and save lives.

All this suggests that the current workforce-automation trend is beneficial not only for the economy, but also for the workers in those jobs—provided that those laborers can develop the new, relatively advanced digital skills required to manage the machines or find other decent-wage jobs. The problem is that few Latinos seem to have these opportunities. According to the Pew report, Latinos are far more likely than whites and blacks to cite automation as the key cause for why their hours or pay have been reduced, or even why they lost a job. While the construction sector has many openings, for example, sometimes those openings are in new areas to which displaced workers can’t travel, as my colleague Alanna Semuels has pointed out in her reporting.

This reality isn’t surprising. In an effort to gauge the impact of automation on different racial groups, the Brookings Institution’s Muro and Whiton assessed the “automation potential”—a measure developed by McKinsey that pertains to how many tasks can be automated using today’s technology—for the 20 occupations in which each racial group is most concentrated. They then compared that to the number of workers in each occupation. Latinos, the researchers found, face the highest automation potential at close to 60 percent, followed by blacks at 50 percent, Asians at almost 40 percent, and whites at roughly 25 percent. Without digital skills, Latinos can be automated out of jobs that extend far beyond the construction and hospitality sectors: According to Brookings, the share of jobs that don’t require workers to have experience beyond basic digital skills—such as knowing how to use Excel and Word—has fallen from 56 to 30 percent between 2002 to 2012.

Proposals to tax robots and automation, among other efforts, seek to curtail the spread of new technologies. But assuming those attempts are unsuccessful, creating more education opportunities that target Latinos could help improve their employment prospects. Entrenched school and housing segregation means that Latinos have far less access than whites to resources that determine their longer-term education and job trajectories, largely by influencing who gets what skills early on—who, for example, gets the schooling that allows her to pursue a computer science degree at MIT and who gets the schooling that all but limits her to a job in housekeeping. Latinos have a higher high-school dropout rate—at 10 percent—than their black, whites, and Asian peers, in large part because they tend to pool their resources together to have one household income, said Jaime Dominguez, a political-science and Latino-studies professor at Northwestern University, alluding to the fact that so many Latino families are low-income. “There is an obligation to work.”

Training and education programs that focus on teaching participants the digital skills they need to thrive in the current economy will have to ensure that those participants can earn while they learn. Certain initiatives, such as Pathways 2 Apprenticeship, focus on African-Americans, Latinos, and other disadvantaged people of color namely because they are more likely to be low-income than whites.  

Automation threatens to exacerbate a pattern in which Latinos are stuck at the bottom of the socioeconomic ladder: In depriving them of jobs—and often pushing them into the devastating cycle of long-term unemployment, the trend will make it increasingly difficult for low-income Latinos to enter the middle class.

Yet few resources have been dedicated to ameliorating the trend’s impact on the demographic’s workforce. The most recent federal-job training program, the Workforce Innovation and Opportunities Act of 2014, which President Obama signed in 2014 in an effort to improve the employment prospects of working Americans, has specific provisions for Indian, Alaska Native, and Native Hawaiian individuals in recognition of those groups’ unique disadvantages. Doubts about the merits of federal job-training initiatives aside, it’s worth noting that the act does little to explicitly target supports for Latinos.