Quick question before the data. Do you know why the federal government built a job training system in the first place?
September is #WorkforceDevelopmentMonth, created by the National Association of Workforce Development Professionals (NAWDP) in 2005. Last year was the twentieth, so this is the twenty-first. Most of us will spend it posting success stories, and we should, because those stories are real. I want to spend mine on a piece of our own history nobody brings up, because it has gotten suddenly, uncomfortably relevant.
On March 15, 1962, President Kennedy signed the Manpower Development and Training Act (MDTA). Its purpose was to train and retrain workers unemployed because of automation and technological change. Yes, you just read that. Sound familiar? The MDTA is arguably the founding document of the modern American workforce system, and machines taking jobs is written into its reason for existing.
Here is the detail that stopped me. The original bill limited retraining to unemployed workers over 30 who were heads of families with five or more years in the labor force. Congress aimed the first federal training dollars at experienced adults, deliberately not at young people. Hold onto that.
Two numbers from this summer
On August 12, Erik Brynjolfsson and colleagues at the Stanford Digital Economy Lab updated their analysis of ADP payroll records covering millions of workers. Employment for 22 to 25 year olds in the most AI-exposed occupations now sits about 19 percent below where it would be if it had kept pace with less-exposed peers, from November 2022 through June 2026. A year earlier that gap was 15 percent. They call it correlation, and they found no economy-wide displacement. What they did find is that the decline comes from reduced hiring, not separations. Companies are not firing young people. They are quietly not starting them.
Closer to my neck of the woods, a report commissioned by Workforce Solutions Rural Capital Area with ARMA, Workforce Solutions Capital Area, and Opportunity Austin projects upwards of 33,000 new semiconductor and advanced manufacturing jobs across twelve Central Texas counties by 2030. Hardest roles to fill? Assembly technicians, orbital precision welders, CNC machinists.
One dataset says young workers cannot get hired. The other says we cannot find enough of them. Both are true at once, and that is not a contradiction. It is the whole plot.
Codified versus tacit
The Stanford team’s explanation is the sharpest I have read. AI substitutes well for codified knowledge, the documented material you write down and teach in a classroom. It complements tacit knowledge, what you pick up through practice, mentorship, and situations that do not match the manual. Sit with that, because it is a curriculum statement disguised as a labor market problem.
Codified knowledge is the cheap half of what we teach. Syllabus, LMS, six-week course, stackable certificate. Tacit is the expensive half. Equipment, lab hours, small ratios, an employer willing to host, and someone with twenty years of scar tissue standing next to someone with none. Guess which half our funding systems have quietly been optimizing for.
Sixty-four years of getting one thing right
Look at what our field did between 1962 and now, because the pattern is remarkably consistent. The Comprehensive Employment and Training Act (CETA) in 1973 pushed decisions out to roughly 470 local prime sponsors. The Job Training Partnership Act (JTPA) of 1982 routed money through governors to Private Industry Councils and required most of it go to actual skill development. The Workforce Investment Act in 1998 built the one-stops, replaced those councils with workforce boards, and handed people Individual Training Accounts so they could choose. WIOA in 2014 made the system demand-driven and imposed common performance measures across programs. (I know we all have strong opinions on WIOA reauthorization and training mandates, I’ll circle back to that soon!)
Texas HB 8 in 2023 is that arc arriving at community college finance, and arguably the most serious version of it anywhere. To count as a credential of value, a typical student has to earn enough within ten years to cover the cost of that education and beat a typical high school graduate’s earnings. Colleges responded, and then some. Between fiscal 2023 and 2025, credential attainment rose about 22 percent and dual credit hit an all-time high, up roughly 20 percent. The model worked so well it outran its own appropriation, which is why the Coordinating Board spent July adjusting formulas. Nobody did anything wrong here. The incentive did what a good incentive does. And yes, there is still tons of debate about this.
What sixty-four years never had to handle
The arc assumed one thing that held for sixty-four years and may not anymore: that we could measure the labor market faster than it changes. Folks, those days are GONE.
Our methods for deciding what a credential is worth runs on a rear-view mirror. Ten-year earnings horizons. Wage record matches that arrive years after graduation. Program approvals built on projections written before the last two model releases. That was reasonable when the entry rung moved on a decade. It just moved four points in about a year.
And notice where it moved. Kennedy’s system was pointed at experienced adults over 30 because that is who automation displaced in 1962. The displacement showing up now is at 22 to 25, among people who do not have a job to be displaced from yet. The problem has migrated to the opposite end of a career, and the entry point is the one part of the pipeline our system was never built to defend.
Here is the part that keeps me up. When budgets tighten and you are counting credentials, the rational move is to produce more of the cheap kind. That is exactly the arbitrage AI is exploiting. I say it as someone who built programs that counted. As a senior executive in higher ed, I knew what a welding lab cost per student versus a business certificate, and which one made the enrollment math work. Nobody in those rooms was cynical. We were responding honestly to what was being measured.
Three things I would put on the table
I am not arguing against outcomes funding, quite the opposite actually. I worked on the advanced manufacturing pipeline for Empire State Development and on semiconductor strategy for the White House Workforce Hub in Columbus, and in both places the challenge (and opportunity) of tighter outcomes tracking was top of mind.
Perhaps we pair the ten-year wage match with a real-time employer hiring signal, so a drifting program gets flagged in eighteen months instead of six years. Build a cost weight that recognizes tacit-knowledge instruction is structurally more expensive per student, rather than letting a credential count reward whatever is cheapest to produce. And ask honestly whether “high demand” and “AI resilient” are the same list, because right now we assume they are.
Central Texas is better set up for this than most regions. The Coordinating Board is iterating in public, our colleges are deep into semiconductor training, and employers have put their hard-to-fill roles in writing. Those welders and machinists are not a nostalgia play. By the Stanford data’s own logic, they are some of the most AI-resilient work in the economy.
So here is my Workforce Development Month question. Our field was founded on the premise that technological displacement is something you retrain your way out of after it happens to you. If it is now happening to people before they are ever hired, what would it take to fund the expensive half like it is the point rather than the overhead?
Twenty-one Septembers in, that feels like the right thing to argue about. I’d love to connect and kick this around with you. In the meantime, hug your local workforce board career coach, college enrollment coach, registered apprenticeship sponsor, and economic development pro. We’re doing amazing work, and you can be proud!
Edgar Padilla is the Founder & Chief Strategist at Padilla Futures Group, a workforce, education, and economic development advisory firm. He is currently embedded with the City of Austin Economic Development Department as a FUSE Executive Fellow, focused on the advanced manufacturing and semiconductor industry and workforce. He is a former senior executive in technical higher education, publicly-elected school board trustee, Deloitte and EY alum, and experienced non-profit, EDO, and education board member. Views are his own.
Originally published on LinkedIn.