The Federal Opportunity Zone tax incentive program is considered by many to be the Trump Administration’s flagship economic development initiative.  It’s a big money program, with more than $108 billion now sitting in Opportunity Zone funds across the US.  The OZ program began in 2017, and was revised and updated in 2025.   

Despite more than five years of operations, the jury is still out on whether opportunity zones work.  In my view, the evidence supporting the OZ program’s effectiveness remains quite weak. And, it seems like I’m not alone as a recent Government Accountability Office (GAO) assessment offers a sober and critical look at the OZ program and its effects.

GAO analysts assessed the state and local impacts of OZ programs, and found that few programs actually track job creation and other impacts from OZ investments.  Not surprisingly, surveyed state and local officials were typically “unsure” of any concrete impacts. Their survey responses noted that they hoped for impacts on poverty, jobs, and development, but that evidence in support of these impacts remained “anecdotal” at best. GAO offers a number of recommendations for improving the program, but it should be a concern to policymakers that one of our largest economic development investment programs still lacks the capacity to track its impacts and its return on investment to taxpayers and to local communities.

While the jury is still out on how AI will affect job markets, warning signs abound and policymakers are searching for ways to ease the transition.  Not surprisingly, expanded job training programs rank at the top of most policy solution sets.  While I applaud this belated appreciation of workforce development programs, a new report from Anthropic cautions that training alone will not solve the AI job transition challenge. 

“An Evidence Review of Workforce Training” offers a detailed policy and literature review of workforce training models and impacts.  The findings focus on stand-alone programs, as opposed to training programs at community colleges or apprenticeships.  The research finds that training programs do help, but their impacts can be relatively small. They tend to have real, but small, effects on overall  employment level (up an average of 1.7%) and incomes (with average increases of only $800).  Sectoral programs, such as cluster or employer-led efforts, can have bigger impacts, but more research is needed on how these programs work.

What can we conclude from this sobering research?  My take is that addressing AI disruptions will require more than simply spending more on current workforce programs.  We will need to rethink programs, repackage them, and consider bringing in new models and approaches.  This will be a tough challenge, but maybe there is a silver lining.  As we deal with AI disruptions, we may finally get serious about building robust and effective workforce development systems that American workers need and deserve.