There are plenty of smart people who want to build with AI. What’s missing is the on ramp.

Every week, I see another post about how hard it is to find “AI talent.” And every week, I see the job descriptions those same companies are posting: entry-level roles asking for years of production AI experience for a technology that has barely been in production that long.

Something does not add up.

Sit with that for a moment. We are, collectively, an industry asking for proof of experience we never created the conditions for anyone to gain. Then we act surprised when the pipeline runs dry.

Meanwhile, I talk to graduate students who understand this technology deeply. They can explain the architecture, tradeoffs, and limits better than many of the people interviewing them. But they still cannot get past a resume screen because they have never “shipped to production” or gained real-world experience building AI systems.

Of course they haven’t. Almost nobody has given them the chance.

The gap nobody’s funding

This is the part of the conversation the industry tends to skip over.

We talk endlessly about how desperate we are for people who can build with AI. We invest heavily in recruiting people who can build with AI. We write think pieces about the cost of not having people who can build with AI. What we don’t do, at anywhere near the scale required, is invest in the infrastructure that turns capable students into production ready builders.

This isn’t a hunch. SignalFire’s analysis of major public tech firms and maturing venture backed startups found a 50% decline in new role starts by people with less than one year of post graduate work experience between 2019 and 2024. The decline held consistently across sales, marketing, engineering, recruiting, operations, design, finance, and legal. Revelio Labs data cited in the same reporting shows U.S. entry level job postings overall have fallen roughly 35% since January 2023. Meanwhile, NACE’s Job Outlook 2026 survey of employers, reported in IEEE Spectrum, confirms what every recent graduate already knows: industry experience and demonstrated proficiencies now sit at the top of what recruiters and hiring managers actually look for.

The result is predictable. A generation of graduates who’ve taken excellent courses, completed thoughtful capstone projects, and read every paper worth reading, only to get filtered out at the first screening question because their resumes don’t include the magic words: deployed, production, real world impact.

This isn’t a talent crisis. It’s a pipeline failure. And it’s one the industry has built for itself.

What we decided to do about it

Last week, we launched the Student Builder Program at Claremont Graduate University, home of the Drucker School of Management and the Center for Information Systems and Technology (CISAT).

It is not an accident that this program launched at the Drucker School. In Managing the Non-Profit Organization: Practices and Principles (HarperCollins, 1990), Peter Drucker argued that nonprofit institutions sit at the center of modern society. What makes them distinctive isn’t their tax status. It’s their work. The actual product of a nonprofit, in Drucker’s framing, is a changed human being: a patient who recovers, a student who learns, a community that becomes more capable than it was before. That framing is woven into the DNA of this program. We’re not training students to optimize ad clicks. We’re training them to build technology for institutions whose product is human change and comes out of the program more capable that they were before. 

The model is straightforward:

Students train on the AISquared platform. The same one we deploy in the field for our largest customers. The thinking is simple: if we want students to develop production-grade skills, they need to learn on production-grade tools.

They earn a certification as AI builders. A credential that signals real capability, backed by a curriculum that mirrors what we’d teach a new hire.

We match them with vetted nonprofits. Organizations doing meaningful work in their communities who couldn’t otherwise afford to bring AI into their operations. These are problems with a deadline, a budget, and someone counting on the answer.

Students deploy real workflows into real environments. Messy data, organizational complexity, the kind of ambiguous requirements that don’t show up in coursework. The work is hard. That’s the point.

They walk away with proof. Portfolio projects, references from nonprofit leaders, demonstrable outcomes. The kind of evidence that changes how a recruiter reads a resume.

And because we want to remove every excuse not to build, we’ve structured the program so the economics work for the people in it. Accepted students get free access to the AISquared enterprise platform. Not a stripped down student trial, not a sandbox version, but the same product we license to the U.S. Department of War and to Fortune 500 enterprises. The value compounds across everyone who participates. Students earn the AISquared Certified UNIFI Associate credential, a deployed portfolio, references from real organizations, and a head start in a job market that’s only getting harder and harder to break into. Universities get a differentiated program their students can’t access anywhere else, plus a pipeline story to tell prospective applicants and accreditors. Nonprofits get AI capability built into their actual operations, not a deck full of recommendations. Everyone leaves with something they couldn’t have built alone.

Two things worth saying out loud

Nonprofits are not charity cases in this story.

They’re some of the most operationally complex organizations in the country. They run lean, manage diverse stakeholders, navigate regulatory environments most startups never touch, and serve populations the private sector consistently overlooks. If a student can build AI that works there, under those constraints, with those data realities, they can build AI that works almost anywhere.

Treating nonprofits as a “training ground” undersells what they actually offer: some of the most rigorous deployment environments a young technologist will ever encounter.

Students are not free labor.

They’re the next generation of builders. They’re the people who will define how this technology gets used over the next two decades. The companies investing in their development right now are the ones who’ll have a talent bench when everyone else is still trying to hire one.

This isn’t an internship program. It’s a talent development pipeline, and the value flows in both directions.

What this means for the people reading

If you’re a student looking for the kind of experience that actually opens doors (not another certificate, not another tutorial, but real deployment work you can point to), apply here.

If you’re a nonprofit leader who has been quoted six figures for AI work and walked away, or who knows your team needs AI capability but hasn’t been able to access it, apply here.

If you’re an educator, a university leader, or a talent leader who wants to bring something like this to your community, my DMs are open. We want to expand this. We can’t do it alone.

What’s next

CGU is the first. It is not the last.

We launched at Claremont because the Drucker School is the right philosophical home for this work, and because we found a community of educators, students, and operators willing to build something with us. But the talent gap doesn’t end at one campus. The pipeline failure I described earlier is industry wide. So is what comes next.

We’re already in conversations with other institutions about bringing the program to their students, and with other nonprofit networks about expanding the pool of organizations students can serve. 

If you lead a graduate program, run a nonprofit coalition, or sit somewhere in between and see a way to plug in, please reach out. The model is built. The question is how fast we can scale it.

The bottom line

The talent is already out there. It always has been.

What’s been missing is the bridge between capable students and the production environments where they can prove themselves. The companies that keep waiting for “qualified AI talent” to magically appear are going to be waiting a long time. The ones that decide to build the bridge themselves will inherit the talent pool.

So, we decided to stop waiting.