As policymakers grapple with how artificial intelligence will reshape work, they face the familiar temptation to double down on training. The logic is that, if AI changes the knowledge workers need, then workforce systems should adapt quickly to teach in-demand skills.
However, this framing assumes the main problem is a shortage of skilled workers, when in many high-paying sectors the bigger problem is a shortage of entry-level opportunities. This dynamic is not unique to AI. After the Great Recession, as digital skills became the new must-have skillset, cities across the country invested in free coding programs built on the premise that if they taught people to code, good jobs would follow. Federal initiatives like the Obama-era “TechHire” poured millions of dollars into these efforts.
The unfortunate truth, however, is that many of these programs struggled or shut down because training workers did not automatically translate into jobs. As many programs found out, access to entry-level positions in the high-wage tech sector proved far narrower than policymakers assumed, and employers were reluctant to hire workers with limited experience and often required extensive screening.
The core problem facing policymakers in the AI age is the same: an incentive mismatch.
Expanding entry-level opportunities creates enormous social value in the form of higher earnings, stronger families, and a more capable workforce. But for employers, hiring and training new workers remains risky and expensive.
If policymakers want to change that, training alone won’t be enough. They will need to reduce employer risk and help share the cost of developing talent, likely through stronger public-private partnerships that make it easier to hire and retain entry-level workers.
The most successful coding programs solved this problem directly. They reduced hiring risk by tightly screening participants, investing heavily in case management, and aligning closely with a small number of willing employers. In effect, they acted as labor-market intermediaries.
That experience offers both a warning and guidance for the AI era. If skill expansion alone could not solve the entry-level job problem during the tech hiring boom, it surely won’t solve it now. As artificial intelligence allows experienced workers to do more with fewer junior workers, the number of true entry points into high-paying careers will shrink even further.
Fortunately, the decade-long experiment in free coding programs also offers several lessons for policymakers on how to keep entry-level pathways open in the age of AI.
One idea that has gained traction among policymakers and workforce experts is expanding apprenticeships for white-collar jobs, which can effectively bridge the gap between training and employment. Take, for example, LaunchCode, where prior to full-time employment, participants enter a six-month apprenticeship with employers, allowing companies to evaluate candidates on a contract basis, thereby reducing the risk of hiring entry-level talent. One might be skeptical and think employers would just churn through contractors, but the conversion rate for LaunchCode is relatively high, with just over 80% of apprentices being offered full-time jobs.
Another promising approach is modeled by the Per Scholas technology training program, which provides ongoing support for its learners after they complete training and enter the workforce. About 80% of their graduates find a job within a year, with 63% entering the tech sector, and alumni benefit from continuing and complimentary training courses, financial planning, and career coaching for up to two years after graduation. This support helps workers succeed early in their careers, and research has shown the turnover rate for Per Scholas graduates is about 22% lower than for other, similar hires, with more than half remaining with the same employer for at least three years.
Finally, workforce systems may need to adopt a more integrated approach to ensure trainees develop the skills employers require. Many free coding programs emphasized close connections with employers through curriculum review, classroom visits, and mock interviews. However, a more effective approach may have been customized job training programs tailored to meet the needs of particular employers. Evaluations of customized job training programs have shown increased employee retention, as well as improved output and team performance for employers.
While evidence shows that apprenticeships, case management, and customized job training can benefit employers, the evidence also shows that most companies won’t make those investments on their own. The AI revolution, with its continued uncertainty and upheaval, will likely only make employers more reluctant to make large investments in entry-level workers.
The public and private sectors will need to work closely together to reduce the risk of hiring and to connect talent with opportunity. If we ask workers to adapt, we have to make sure good jobs are there for them when they do.
Kathleen Bolter is the principal for the Policies for Place initiative, which studies community strategies for good jobs, at the W.E. Upjohn Institute for Employment Research.