For decades, the first job has been where most careers really start. You learn the routine tasks, make mistakes where they are cheap, and slowly build the experience that the next employer asks for. Generative AI is now taking over many of those routine tasks, and that raises an uncomfortable question: if the first rung of the ladder gets thinner, how does anyone climb it? HP and the HP Foundation have put money and training behind five organizations that are trying to answer exactly that, and the results of their Pitch Fest in New York offer a useful look at what practical solutions might look like.
Why the first job matters so much
It is easy to treat entry-level work as simple work that anyone could do. In practice, it is how organizations pass skills to the next generation. A junior analyst who spends months cleaning data learns what good data looks like. A new developer who fixes small bugs learns how a real codebase behaves. A junior writer who drafts routine copy learns what an editor will cut.
That apprenticeship function is the part AI puts at risk. When a tool can produce a first draft, a basic script or a preliminary summary in seconds, the task that once taught a beginner may no longer be handed to one. The work still gets done, but the learning that came with it disappears, and employers keep asking for experience from people who have had fewer places to get it. That is the catch at the center of the debate, and it is why the question is bigger than whether a given job exists today.
What the data shows so far
The evidence is still being argued over, but the direction is hard to ignore. The Stanford Digital Economy Lab, working from payroll data, found that employment among 22- to 25-year-olds in the most AI-exposed jobs was about 19 percent below the expected trend in its latest update. The researchers say the shortfall comes mostly from companies hiring fewer young people, not from firing those already in the job.
Indeed’s Hiring Lab sees the same pattern in job postings. It reports that the share of entry-level roles among postings in the most AI-exposed occupations fell from 29 percent to 10 percent between 2021 and 2026. The New York Fed puts underemployment among recent college graduates at 42 percent in the second quarter of 2026, with their unemployment rate at 5.6 percent.
Not every study points the same way. A working paper from the CESifo research institute found that unemployment among recent graduates this summer was not unusually high compared with the previous four years. A fair reading is that entry-level work is narrowing and changing in composition, that AI is probably one cause among several, and that the effect could grow if companies keep expanding their use of it.
What the accelerator offers
The Future of Work Accelerator is a joint program of HP Inc. and the HP Foundation. Each selected organization receives $100,000 in funding, $100,000 worth of HP technology and solutions, and a six-month virtual program with workshops, executive coaching and sessions with HP volunteers. HP runs the training with two partners, MIT Solve and Global Impact Advisors.
This is the fifth year, and it looks different from the first four. Earlier cohorts were nonprofits working in countries such as Greece, Indonesia, Nigeria and Spain. The 2026 edition is based in the United States, and for the first time it is open to mission-driven for-profit companies as well as nonprofits. HP says 570 organizations applied, with ideas ranging from AI-powered hiring platforms to gamified career planning. It aimed the program at the people it considers most exposed to economic shocks: young workers, on-demand workers and small businesses.
HP frames the focus with a number from the National Skills Coalition: 92 percent of jobs in the United States now require digital skills. Since 2022, it says, the accelerator has supported 35 nonprofits in 13 countries and helped reach 11.3 million people.
Extern: practice the work before the job
The winner of the Pitch Fest was Extern, and its approach is the most direct answer to the apprenticeship problem. The company builds the infrastructure for externships, which are real, learning-based projects that students complete on behalf of employers such as Amazon, TikTok, Canva, Pfizer, PwC and Home Depot. Students get something to put on a résumé that goes beyond a course grade, and companies get a way to see how a candidate actually works.
Extern says more than 70,000 learners have taken part so far. With the accelerator’s support, it plans to scale its AI infrastructure to serve over 100,000 learners within two years. The pitch is that if the junior job is shrinking, you can recreate part of its learning value in a structured project, at a scale that individual internships rarely reach, and without requiring a student to live near a big employer.
Four other routes into work
The other finalists tackled different parts of the same problem. Millennium Campus Network runs the Millennium Fellowship, a program that trains young civic innovators to work on community challenges and prepare for careers in social impact. Its bet is that collaboration, leadership and judgment stay valuable even as routine tasks are automated.
PHIND Work, based in Colorado, is a mobile platform that uses location matching to connect people with jobs, training and resources in construction, the trades and clean energy. These are fields where much of the work is physical and tied to a place, and where finding the right entry point can be as hard as learning the skill.
Skillionaire Games, from the company skillsgapp, uses career exploration video games to introduce young people in underserved communities to possible paths. And SkillUp is a nonprofit coalition of 150 partners that helps workers from under-resourced communities build skills for in-demand jobs, aimed at people who do not have, and may not want, a four-year degree.
It is early to celebrate
A pitch event is not a result. Five organizations cannot repair a labor market, and the figures above show how large the problem is. The accelerator itself acknowledges this by asking participants to report their reach over several years, at a baseline, then after six months, eighteen months and two and a half years. The program runs from May to October this year, followed by an alumni phase that begins in November. The real test will come when those later numbers arrive.
It is also worth remembering that part of the package is HP’s own hardware. That is a natural fit for a technology company, and the equipment is useful to small organizations, but it remains a corporate program with business interests behind it. HP has set a goal of reaching 150 million people through its social impact programs by 2030, and it counts 65 million through 2024, which shows the accelerator is one piece of a much larger effort.
What employers can do
The most useful lesson from the five finalists is practical rather than philosophical. If AI takes over the tasks that used to train beginners, someone has to build another way to get that experience. Externships, trades pipelines, guided career exploration and retraining are all attempts to do that from outside the company.
Employers can do their part from inside. They can keep a reasonable number of true entry-level roles, even when AI could do part of the work, so that today’s beginners become tomorrow’s experienced staff. They can open real projects to students and career changers, as Extern’s partners do. And they can review job postings that ask for years of experience in roles labeled as entry level, which makes the first rung harder to reach than it needs to be. Companies that sell AI tools are in a particularly good position to lead, since the technology they promote is part of the reason the ladder is changing.
What this means
For anyone starting a career, the message from this program is that the traditional path is no longer the only one, and that proof of real work matters more than ever. For companies, it is a reminder that experienced workers do not appear from nowhere, and that training the next generation is now something that has to be done on purpose. Whether these five organizations can scale their approaches is still an open question, but the problem they are working on is not going away.
