Employment is America's load-bearing institution, AI will test it

Engineers have a rule: never design a system with a single point of failure. A bridge should not collapse because one cable snaps. A power grid should not go dark because one transformer fails. Resilient systems distribute risk, build in redundancy and assume shocks will come. For decades, the United States has done the opposite. We have built our economy around a single load-bearing institution: employment. For most people, a job is more than a paycheck. It's how millions of Americans get health insurance, save for retirement, arrange childcare, establish credit, access paid leave and, for some workers, maintain legal immigration status. This Labor Day, as we recognize the contributions workers make to our country, we should also confront just how much of American life rests on the stability and quality of their jobs. We have attached an extraordinary amount of economic security to this one institution, based on the assumption that employment will remain stable and insulated from major shocks. Artificial intelligence could very well test that assumption on a scale we are not prepared for. Technology has always reshaped work. The Industrial Revolution displaced artisans. Mechanization transformed agriculture. The internet made huge amounts of information instantly accessible, transforming the media and news industries. Our economy eventually adapted, though not without costs. People lost work; jobs were displaced or relocated; wages were cut. Luckily, those costs were somewhat contained as they were concentrated in particular industries or regions. The speed and breadth of AI today poses a unique threat to working people across many industries. And the concern is not just that AI could eliminate jobs. It could degrade job quality, and make good jobs harder and harder to find. We have already seen a version of this in the gig economy, where algorithms can influence pay, assign work, track performance and surveil workers. Now imagine that model spreading across industries and occupations—from software and finance to customer service, retail, administration, health care and the creative professions. Workers in these fields do not have to lose their jobs for their economic security to erode. A once-stable job with reliable pay, benefits and a path forward can become precarious work with fewer protections and less security. AI is moving quickly through American workplaces. In the second quarter of 2026, more than 45 percent of employed American adults reported using generative AI for their jobs, up from 35 percent one year earlier. And while it's still too early to know whether AI will result in widespread job displacement, employment among 22- to 25-year-olds in highly AI-exposed occupations stood about 19 percent below where it would be if it had kept pace with peers in less exposed occupations. This is why it is foolish to wait for mass unemployment to hit before we take action. When engineers see a bridge's infrastructure is experiencing stress, they don't wait for it to collapse before doing something. They fix the weakened joint or overloaded beam. We need to think about our economy the same way, and we are seeing the stress. Fewer entry-level positions for young people are available, more jobs are being contracted out, hours are becoming more unpredictable and pay is not keeping up with the cost of living. Families are having a harder time saving, buying homes, planning for children and weathering emergencies. Notably, that strain will not be distributed evenly. Women make up 83 percent of workers in the nation's 15 most AI-vulnerable occupations, and women of color account for more than 30 percent. AI has the dual risk of compounding preexisting inequities. None of this is inevitable. If we can recognize these warning signs, we can make the needed reinforcements to hold our economy together. That starts with modernizing unemployment insurance for today's labor market. Our system was built for an entirely different era when careers were more stable and layoffs more predictable. Today, millions move among part-time work, contract jobs, caregiving responsibilities and periods outside the workforce, yet too many remain excluded from the safety net designed to help them weather transitions. Workers also need high-quality skills upgrading, workforce development, job training and apprenticeship programs before displacement becomes permanent. And workers need stronger collective power in their workplaces. When employers introduce AI systems, workers deserve transparency and a meaningful voice in how those systems are used. California is beginning to show what proactive preparation can look like. A May executive order directs state agencies to track early warning signs of AI-driven job disruption, consider updates to the state's WARN Act and worker safety net programs, and explore ownership models that could allow workers to share in productivity gains. None of this requires resisting innovation. But we need technological progress to expand opportunity, not transfer more risk onto workers with the least room to absorb it. Engineers understand that resilience is measured by how a system performs under stress. AI may become the first true stress test of an economy that asks employment to do too much. If work is expected to provide health care, retirement security, career mobility, family stability and access to opportunity, then every disruption to work puts pressure on all of those systems at once. We must recognize that the foundations of economic security rest entirely on employment and job quality. As AI puts that structure under strain, we need to build the supports that keep shocks from becoming crises. The future of work and American workers depends on it. Rebecca Dixon is president and CEO of the National Employment Law Project. The views expressed in this article are the writer's own.