from
It’s often said that there’s no first anything. Whenever anything is claimed as “the first” -- e.g., first in flight, first rock-and-roll song, first athleisure suit – it’s possible to point to an earlier example. The same is true of jobs. What was your first job? My immediate answer is scooping ice cream at Baskin Robbins – my first paycheck. But that wasn’t the first time someone paid me to do work. A year earlier, my mother, a sociology professor, somehow convinced the administration to put her in charge of the community college’s new “Staff Computer Centre” and nepotistically paid me to teach faculty and staff to use MacWrite and MacPaint. And even before that, a friend of mine with a paper route went away for winter break. So I filled in for him – delivering the Toronto Sun, a British-style tabloid with a half-naked “Sunshine girl” on page 3 – and received generous Christmas tips from perverts.
But when I think about my first job, I don’t think about any of the above. Because a month after I started at B&R, I got a much better job just up the street at Oliver’s Bakery Restaurant. Oliver’s wasn’t fancy, but they baked their own breads, croissants, and cakes and ground their own coffee. The place was cozy and smelled wonderful. I worked there as a busboy and server throughout high school despite the fact that the backroom smelled less wonderful. Oliver’s backroom featured a staff table full of dirty dishes and empty beer bottles, lockers where our food-coated black sneakers festered, and never-cleaned floors dotted with glue traps for mice. After especially hectic dinner shifts, I’d have nightmares about chaotic moments in my section, and especially about that backroom.
Although Spiderman: Brand New Day and The Odyssey were bigger, no summer film was as profitable as Backrooms. Made for $10M by 21-year-old creator Kane Parsons based on horror content posted to his YouTube channel – oddly categorized as ” creepypasta” – Backrooms got Gen Z off their phones and out of their parents’ houses, generating nearly $400M at the box office. Inspired by an early aughts online image of a Wisconsin store with lots of partitions and fake inner walls, the film follows a furniture store owner who finds a secret doorway that leads him to an endless series of dimly lit, nondescript rooms – strange, flawed, workerless spaces that evoke the interminable doomscrolling and depersonalization of digital life. The backrooms are empty save for a monster and faulty copies of people who move as unnaturally as SBA Administrator Kelly Loeffler. While prior generations were alienated by wars (boomers) or their parents (Gen X), while Millennials are only alienated if their favorite brunch spot runs out of avocado toast, and while I’m only alienated by how we’re suddenly using the word compute (when exactly did compute go from a verb to a noun?), Gen Z alienation stems from not being able to find their way in or out.
Which makes sense. Because far too many find themselves in a liminal space, neither in school nor a career. Needing to make a living in an era of digital transformation and AI, they’re having a horrific time finding their way into and around work. So it’s not just education and training wonks who think workforce is the defining issue of our time. The success of Backrooms demonstrates Gen Z thinks so, too.
It’s happening everywhere. Young people around the world are angry at how hard it’s become to launch careers. This summer an entire protest movement arose in India – the world’s largest Gen Z population – around entrance exams for medical school, one of the few remaining sure bets. In the UK, youth unemployment is over 16%; it’s 15% across the EU and over 25% in Sweden. Back home, nearly 400,000 young Americans stopped looking for work in June alone. So it’s urgent that we identify what actually works to help young people get good first jobs.
The fact that we haven’t figured this out yet is both bananas and bleak considering what AI is about to do to the labor market, as Bill Gates warned this week. The reason is simple. State and local workforce agencies have never been held accountable for anything more than spending their budgets, enrolling job seekers in training programs, and – infrequently and at best – placement rates without regard to job quality or career path. Nonprofit workforce organizations are accountable only to their boards and funders – often the same public agencies – who don’t ask hard questions. And forget about colleges, which dominate post-high-school labor market investment (over 95% wallet share of government spending). By September, they have no idea what their May graduates are doing, nor has anyone asked them to seriously answer that question.
Earlier this month, America’s authoritative source for rural news, the Daily Yonder, ran an article on a new Virginia workforce development program. The headline: This is the Model. There’s a hunger to answer the question of what works before countless jobs are transformed by AI and before millions of Gen Z job seekers who earned degrees with the expectation of following their parents’ white collar path are informed they need to recalibrate and work with their hands.
Here’s what we do know: apprenticeships work well. They’re the gold standard because the search for employment doesn’t involve an endless series of creepy, nondescript, flawed rooms. Apprenticeships are jobs from day 1 with built-in training. But decades of wrongheaded policy and underinvestment left us with virtually no apprenticeship infrastructure outside the building trades. And despite effort, advocacy, and unprecedented momentum, it remains a high bar to convince U.S. employers to engage with young, untrained, unproductive workers. So apprenticeships are unlikely to scale quickly enough to avoid anger, protests, and upheaval.
What scales? Training does. In theory any number of job seekers can be put through an education or training program. That’s the beauty and simplicity of pushing students into classrooms and sitting them in front of instructors. But as shown in the comprehensive training takedown released by Anthropic earlier this month, they don’t work. It’s why keen observers like Brookings’ Molly Kinder recognize that “we can’t retrain our way out of AI’s economic disruption,” why I’ve taken to calling standard-issue programs “train-and-pray,” and why so many workforce prayers go unanswered.
The answer lies between apprenticeship perfection and train-and-pray scale. We desperately need a scalable training-first model that’s as proximate as possible to employment opportunities, employers, and actual employment. That can only mean work-based learning: real work experience completed during or connected with a training, certificate, or degree program. Work-based learning can take many forms: rotations, clinics, co-ops, internships, short projects. It can be integrated into coursework or independent. But what all forms require are bona fide employers, which regulate scale. Such programs can enroll only as many as can fit within the constraint of available work-based learning opportunities.
Which creates a temptation to simulate real work in order gain greater scale. That’s what Mark Cuban predicts: AI-driven simulations e.g.,“junior lawyer rehearsing a deposition, a doctor responding to an AI patient, or an electrician diagnosing a simulated fault.” But as long as we’re talking about real employers (not AI or simulated employers), we’ll need evidence that real employers value simulated work. Until that’s clear and convincing, we should be doing everything we can to scale real work-based learning, which could include coming up with creative and tech-driven ways to allow hundreds or thousands of students to do the same real work for the same real employer.
Having established these parameters, we’re trying to solve for: (1) which skills to train on; (2) how to deliver training; (3) which form of work-based learning; (4) identifying/convincing employers. It’s a multivariate equation that’s complicated by a conflict. Once a training-first program is stood up with resources for developing and refreshing curriculum, delivery, and assessment, it’s effectively permanent. But neither labor market demand nor employer willingness to participate in work-based learning schemes is permanent – particularly for specific jobs in specific sectors. So any solution needs to negotiate that different rhythm.
Given this complexity, plausible answers depend on data. It’s imperative that workforce resource and design decisions be informed by labor market data, enrollment data, and enhanced wage records – all integrated. It’s equally essential that individual job seekers have access to the same, as well as data relating to current and potential public benefits (because life gets in the way of training – even in the way of work-based learning). Employers also benefit from data, which can convince them to provide work-based learning opportunities. This is why emerging talent marketplaces and labor exchanges like FutureFit AI are so important. Without these platforms, no local or state entity can hope to figure out what works or where investment should be directed. And job seekers are likely to find themselves lost in the backrooms.
***
Some states are making progress on both fronts: amplifying earn-and-learn and implementing talent marketplaces. Virginia directed all four-year colleges and universities to integrate work-based learning into degree programs and Ohio has begun funding internships and co-ops, although neither has begun shifting resources out of train-and-pray programs. But who can blame them? We don’t have data to show what works. But we do have progress. Connecticut is reporting “unparalleled results” from its new FutureFit AI-powered talent marketplace, Career ConneCT: of 7,500 job seekers served, 80% have completed upskilling programs and 80% have been placed. Massachusetts is implementing a similar statewide labor exchange. Alabama has already established Alabama Works. Most states are nowhere, but no state thinks its current workforce system is doing a good job, let alone up to the task of what comes next.
Enter former U.S. Secretary of Commerce and Rhode Island Governor Gina Raimondo. Her new nonprofit RAISE US is the most important effort yet to answer the question of what works. With over $500M in commitments from AI and philanthropy giants, RAISE US is explicitly focused on earn-and-learn models and working with states to figure out what workforce infrastructure to build and scale. The organization touts a policy lab to evaluate what works – the first serious attempt to answer the question no one’s been accountable for asking.
Secretary Raimundo has figured out that workforce development needs a much tighter connection to jobs and employers; we can no longer afford to spend money on training silos. As she wrote earlier this year in the New York Times:
The country needs a modern apprenticeship system that allows workers to earn while they learn, which many European countries have embraced… Skeptics will argue that we’ve tried workforce reform and it hasn’t worked, that the landscape for workforce development is littered with underperforming, small-scale training initiatives. They aren’t wrong. But history shows that real change comes in times of crisis… A.I.-driven mass unemployment is a potential crisis on the horizon. This country cannot withstand the kind of economic shock I see coming. Without solutions, America’s anxiety will become rage — and political backlash will follow, targeting companies that make A.I., businesses that deploy it and politicians who back it.
***
Because we don’t know what works, we underinvest. Total annual federal workforce (WIOA) funding is less than $4 billion, or just over $20 per worker. Figuring out which work-based learning models work for which sector, roles, and geography is key to increasing investment, or shifting investment from other programs. Because there are oceans of spending that could be connected or redirected if we had good answers, including Medicaid ($971 billion) and SNAP ($107 billion) – both with new work requirements – economic development (maybe $100 billion), Workforce Pell (an exciting new program that should scale to several hundred million a year, but one that needs to mandate work-based learning) and – of course – federal and state spending on colleges and universities ($400 billion), including $1.2 billion in Federal Work-Study funding that’s mostly wasted on unrelated, menial on-campus jobs. So it’s easy to see how, without increasing overall spending, we can afford work-based learning and talent marketplaces for every American lost in the backrooms.
AI-driven job loss without proven, investable models for career launch and redeployment will leave a generation wandering an endless series of dimly lit, nondescript spaces with no way out. And if our failure to deal with manufacturing job loss gave us Donald Trump, what kind of creepy, faulty copy of a president is around the corner?