Comments
OFF THE GRID - The “hard-to-employ” is a term that came into use in the 1960s and 1970s to describe groups of workers seen as outside the work world: long term welfare recipients, ex-offenders, out-of-school youth, persons in recovery. Within these groups the workers differed greatly in skills, behaviors and work motivations. Within each group, though, existed a significant segment of workers disconnected from jobs. In March 1968, President Lyndon Johnson, announcing a massive jobs initiative, would reference the “hard-core unemployed."
For over six decades America’s vast workforce system has sought to bring these workers into steady employment—for their own benefit, the benefit of their families and the benefit of the broader society. Success has been limited. Even when job placements have been achieved, they often have been lost: the workers let go or leaving on their own.
The rapid take up of Artificial Intelligence (AI) in workplaces is widely seen today as a threat to employment across the job market. Yet, for the workforce system, AI represents potential breakthroughs in improving employment levels, across the board and especially for the hard-to-employ.
Applications of AI in the workforce system are emerging, and more are likely to launch in 2026. AI holds promise of breakthroughs in assisting the hard-to-employ to navigate the job world and find and hold jobs, in resetting job training and placement programs—which haven’t fundamentally changed in decades, and in designing new forms of employment for workers with more severe mental health challenges and developmental disabilities.
AI is often portrayed as hastening greater unemployment and poverty. Instead, it may offer solutions.
Jobs, Not Guaranteed Income
Whether it does produce solutions will start with whether policymakers can resist the siren songs of the guaranteed income/universal basic income proponents. The rise of AI has given new momentum to guaranteed income and universal basic income initiatives that emerged in the 2010s.
The guaranteed income initiatives never were financially viable; the ambitious policy entrepreneurs, journalists and politicians pushing them could never explain how they would be paid for. These initiatives downplayed the roles that jobs play in finding meaning and structure—especially for the hard-to-employ.
Two essays this past fortnight remind us of these roles. “As artificial intelligence advances, some are beginning to welcome a future without work,” notes Manhattan Institute Senior Fellow Rob Henderson. “But giving everyone a universal basic income won’t reveal most people’s inner Mozarts. It will make them profoundly unhappy.”
CEO: C-suite news, analysis, and advice for top decision makers right to your inbox.
A small share of people, unshackled from jobs, use their free time to create build and explore, Henderson argues. But for most, that isn’t what occurs, and the lack of structure is often disorienting and gradually a source of unhappiness.
Jason Riley looks at experiments with basic income payments that have been undertaken between 2017-2025, some of them funded by the tech aristocracy. “Tech billionaires…seem to believe that the work ethic is an expendable feature of a free market society,” Riley explains. “It isn’t. If basic income advocates succeed, one of AI’s legacies might be forcing future generations to earn these painful lessons all over again.”
Most job counselors and other practitioners in America’s vast public workforce system agree on the power of the job—it is this power that attracted them to their field. But they struggle to access this power in increasingly competitive labor markets and with workers who aren’t fitting in.
AI and the Hard-to-Employ (1): Advances in Navigating the Job World
In theory, when a welfare recipient, ex-offender or out-of-school youth comes for employment assistance, the job counselor will develop an individualized employment plan. The plan is meant to identify occupations and local employers to pursue. The job counselor helps develop a resume, provides interview practice, helps the job seeker address transportation or other supportive services.
In practice, job counselors are often overwhelmed with caseloads of over 70, and the services provided are often cursory. Moreover, service quality is highly uneven. Younger job counselors, or those new to the field, lack experience with the job search processes, have not honed the techniques of employer or job seeker engagement.
Enter AI.
As Local Workforce Boards, American Job Centers, and state Employment Departments are slowly discovering, AI can a free job counselor from the administrative tasks detailing and tracking services, giving them more time to spend with job seekers.
More importantly, AI enables the job counselor to tap into a knowledge base on hiring dynamics far beyond their individual knowledge base. It enables them to call up information on hiring in real time--no longer depending on government job data that can be weeks or months out of date. It enables them to implement the employment plan with the other government offices that the job seeker is often connected with: welfare, probation, corrections, behavioral health, or developmental disabilities.
John Colborn, executive director of Apprenticeships for America, notes that service delivery protocols for the hard-to-employ have barely evolved in decades, and levels of services have actually declined. AI holds promise of both reviving the intensive one-to-one assistance, and sharpening the quality of services. Apprenticeship programs have far higher placement and retention success than other training programs (the worker is employed from day one of training). But local apprenticeship programs too must contend with the employment barriers among the hard-to-employ groups, and they too are testing AI uses.
Javier Romero oversees the hundreds of job counselors with California’s Employment Development Department. Romero emphasizes to his staff that they must be problem solvers, drawing on resources to address the frequent mental health challenges, chaotic personal lives, and learning disabilities that job seekers bring. Romero and his staff are now studying how AI can improve the “wrap around” supports to bolster job retention.
AI and the Hard to Employ (2): Advances in Designing Job Training and Placement Strategies
State and local workforce boards across the country are tasked with deciding which services to fund, for which groups of workers. To do so, they have looked to the outcome data compiled under the Workforce Investment and Opportunity Act (WIOA) as well as the evaluations by the leading workforce research centers-- MDRC, the American Enterprise Institute, Mathematica, Burning Glass Institute.
Blake Konczal, director of the Fresno Regional Workforce Development Board, sees in AI the ability to take evaluation to a higher level and also to individualize it to specific geographies and specific workers. Konczal explains, “What excites me about AI is that it can help us see much more clearly what’s working, for which workers, and in which parts of a region like Fresno. That means we can adjust programs faster and respond to real employer demand instead of relying only on backward-looking data. Used well, AI doesn’t replace human judgment, it gives local leaders better information to make smarter decisions for the people we serve.”
Workforce groups are reaching out to partner with state and county level mental health departments. The goal: address the mental health conditions and behaviors among the hard-to-employ that undermine employment. Initiatives of cognitive behavioral therapy (CBT), group and individual, have been added to job preparation and retention programs for long time welfare recipients, ex-offenders, justice-involved youth.
Stanford University professor and practicing psychologist Joe Bankman has been studying these CBT initiatives. Bankman notes that “AI dramatically simplifies the task of compiling and evaluating outcome studies. A few researchers with AI will be able to do the work of the many researchers this field needs, but cannot afford to hire."
Further, Bankman explains, “AI could also be used to design more individualized CBT plans. It could help identify the obstacles that are getting in the way of employment for a particular worker, and programs that are most effective in removing those obstacles.”
AI and the Hard to Employ (3): Advances in Job Creation
Beyond job training and placement strategies, the workforce system has tested direct job creation in two forms for the hard-to-employ: transitional jobs as a means of gaining work experience, and ongoing employment for workers with more severe mental health and behavioral conditions, as well as workers with developmental disabilities.
Transitional jobs are utilized today as part of welfare-to-work programs for welfare recipients and re-entry programs for ex-offenders. Outcome data are collected, notably data on movement and retention into ongoing employment. Here, as with the job training and placement strategies, AI offers to bring the analysis and evaluation to a more sophisticated level: tailoring transitional jobs to specific populations and specific workers.
AI’s use in program evaluation and design is likely to play even a greater role with subsidized employment for workers with severe behavioral conditions and/or developmental disabilities. In my experience, for these populations, a job-- somewhere to go every day, a role in the economy, a structure to the day—is often more important than for the general population. Yet, the program opportunities still are very few, and even shrinking in numbers as wage subsidies are pulled back.
The Hard-to-Employ as “Canary-in-the-Coal Mine” for AI Utilization
Whether AI can achieve breakthroughs in employment for the hard-to-employ remains to be seen. Workforce practitioners, such as Colborn, Romero and Konczal will tell you that they are in the early stages of AI utilization. Professor Bankman will say the same about AI use in behavioral health interventions.
But if AI is able to achieve advances in jobs for the hard-to-employ, its use may expand. Professor Bankman describes the hard-to-employ as the “canary-in-coal mine.” The design and service breakthroughs that AI provides for the hard-to-employ could be relevant to broader segments of workers.
AI is usually portrayed as the cause of an unemployment future. Instead, it could be part of a solution.
(Michael Bernick is the former Director of California’s Employment Development Department and previously served eight years on the BART Board. He is currently employment counsel at Duane Morris LLP, a Milken Institute Fellow, and a Fellow at the Burning Glass Institute. A leading voice on workforce issues, Bernick focuses on employment strategies for neurodiverse populations. His latest book is The Autism Full Employment Act. He is a regular contributor to CityWatchLA.com.)
