Licensing the future: Why the ‘Make AI Work for Americans Act’ misses the mark

Senator Mark Kelly’s (D-AZ) “Make AI Work for Americans Act” was introduced recently to much fanfare from representatives of organized labor, activists, and even such prominent economists as Daron Acemoglu and Erik Brynjolfsson. The bill expands unemployment benefits, workforce development programs, and job retraining.

While ambitious in its goal of helping the United States adapt to the changes wrought by AI, it does so without addressing the government-imposed barriers that impede letting people change jobs in the first place, chiefly occupational licensing. In fact, the bill would implement policies that reinforce these artificial barriers.

For instance, the bill’s new national service program would make participants eligible for help with “credentialing costs, licensing fees, exam fees” in fields that rely on human judgment and relationship-based work, and counts “obtaining licensure” as a qualifying outcome. In the full text, licensure appears only as something to pay for, never as something to reform. This is a problem given the high time and resource costs of many licenses. And this is no niche issue: Roughly 22 percent of U.S. workers need a license to work. 

Importantly, these barriers do not just affect the skilled trades, but also “knowledge-based” professions. Accountants, lawyers, physicians, and architects all face licensing requirements. Yet it is these professions where the effects of AI are likely to be felt. Despite the earlier prognostications of some industry executives that AI would create mass unemployment among white-collar professionals, the picture that has emerged is more complicated.

Indeed, while economists have failed to surface clear evidence of AI disrupting the labor market, some evidence has accumulated suggesting that entry-level job opportunities have shrunk in some AI-exposed occupations; however, this remains contested. Conversely, analysts have also found that in some professions, AI exposure increases output and employment. So far, AI has not produced mass job displacement.

Whatever AI’s long-term effects, workers will need to adjust in some capacity, whether to new occupations, new tasks, or new companies. Occupational licensing acts like a tax on each of these solutions. Licensed workers are 24 percent less likely to switch occupations than unlicensed workers, and licensing alone can account for around eight percent of the decline in occupational mobility over the past two decades. Workers in jobs with state-specific licensing requirements also migrate at a rate 36 percent lower than workers in other roles. These license restrictions do not just harm workers, but consumers as well. Licensing raises service prices by as much as 16 percent.  

The bill’s answer is to subsidize these job barriers rather than lower them; if anything, this will only incentivize states to raise their fees and costs. It also does nothing to address the months or years of mandated training and the state-by-state requirements that strand workers who relocated and want a better job. Also, the proposed law doesn’t tackle reforming the boards that decide licensure requirements—boards which are often controlled by established businesses who want to keep out the competition. 

In fact, it may entrench them further since the federally subsidized licensee is another fee paid into the system. And since this proposed program targets jobs that rely on human judgment, it will steer workers toward more onerous (in terms of licensing) professions such as health and personal care. This presupposes that governments can identify and channel workers toward the jobs of the future, a tenuous assumption at best. Government officials and, indeed, private sector managers, are poorly positioned to know what skills and jobs the next swell of creative destruction will introduce or remove from the economy.

Ironically, since licensing often creates artificial shortages, demand for AI to make up the difference will increase. But AI’s impact on the labor market will depend on whether the law allows workers to adapt and choose the next best steps for them. Uncertainty about AI’s effects reinforces the need for labor market flexibility by removing the barriers to entry and exit that slow adjustment. 

For workers in occupations where AI can substitute human effort, fewer licensing barriers ease the search and transition into new roles. Conversely, in those occupations where AI enhances demand for human labor, reducing barriers will help capture the resulting productivity gains.

In Senator Kelly’s own state in 2019, Arizona became the first in the nation to have universal licensing, where out-of-state licensing was recognized as valid for the purposes of getting a job in Arizona. Since then, over 8,000 professionals have been licensed under this law. If Congress wants a workforce fund, it should tie the money to what his home state pioneered: letting a license earned elsewhere count. Otherwise, the bill risks licensing the future instead of preparing for it.