Last week I stood in front of a classroom at Washington State University and looked out at a room full of students who had done everything right. They had studied hard, were earning their degrees, built their portfolios, and showed up with genuine enthusiasm for careers in technology. They wanted to know what the industry looked like on the other side of graduation.  I know the bridge has collapsed for many entry level jobs in tech and I didn’t have a comfortable answer for them.  I admitted this to them but also admitted it was the first time I’ve felt this way in almost 3 decades.

The truth is that the entry-level technology job, the role that has served for decades as the on ramp to expertise, the proving ground where raw talent becomes seasoned skill, is quietly disappearing. Not because those students aren't capable and not because the work doesn't exist, but because somewhere in the boardrooms of American enterprise, a calculation was made: AI can do the work of junior employees cheaply enough that hiring them is no longer necessary.

That calculation is shortsighted in a way that should alarm every person who depends on technology to work well, which is to say, all of us.

The Hollowing of the Middle

For generations, the technology industry operated on a recognizable apprenticeship model, even if it was never called that. A new hire spent their first years doing unglamorous work, including junior support work, writing unit tests, fixing bugs, answering support tickets, running deployments at odd hours. The work was often tedious, but it was also essential training. Through that work, a junior developer or database administrator learned not just syntax and tooling, but judgment: when to escalate, when to push back, how production systems fail under pressure, and how organizations actually function beneath their org charts.

That pipeline is being compressed. Companies that would once have large teams, including senior engineers, mid-level, and juniors, are now hiring just senior engineers and handing them an AI assistant. The pitch from vendors is seductive: AI can handle the boilerplate, the documentation, the first-pass code review, and the ticket triage. Let your senior people focus on what only senior people do best, which is the advanced demands.

What this framing ignores is that senior people became senior people by doing the boilerplate, the documentation, the first-pass code reviews. You cannot skip the apprenticeship and arrive at mastery.  My husband has always said, “It takes 10,000 hours to become an expert in a field” yet we’re asking the next generation to somehow replace the experts without that time investment.

The Burn is Already Happening

As head of PDXWIT (Portland Women in Tech) I hear of layoffs and the flattening of teams. Organizations that eliminated entry-level roles to cut costs, often to invest in AI, are now watching their remaining senior staff buckle under the weight of expectations that were previously distributed across entire teams. AI tools help at the margins, but they do not replace the judgment, institutional knowledge, and human communication that senior engineers carry. Those engineers are burning out, some are leaving the industry entirely, and others are quietly coasting, or doing less because the expectation to do more is unsustainable.

This is not a technology problem. It is a workforce strategy problem that technology is being asked to mask.

The irony is particularly cruel for women and underrepresented groups in tech, who have historically relied on entry-level positions as the foot in the door that more established networks could not always provide. When those positions disappear, the ladder doesn't just get harder to climb and for many, it gets pulled up entirely.

The Catch-22 We’re Just Starting to Whisper About

Here is the deeper problem, and the one that should give even the most enthusiastic AI advocates pause.

The AI systems that are replacing junior workers did not emerge from nothing. They were trained in decades of human-generated work: code committed by junior developers learning on the job, documentation written by entry-level technical writers, support transcripts from help desk analysts in their first year. Every iteration of AI that handles "routine" technical work is drawn on a reservoir of human expertise that was built, painstakingly, by people doing exactly the jobs we are now eliminating.

What happens when that reservoir stops being replenished?

If the next generation of technologists never writes the messy first-draft code, never works through the debugging sessions that teach you how systems actually fail, never produces the imperfect-but-human documentation, then the training data for the next generation of AI grows thinner, more repetitive, and more derivative. AI trained predominantly on AI-generated output is already showing signs of what researchers call model collapse: a gradual degradation in quality and diversity as the signal-to-noise ratio inverts.

We are, in the language of farming, eating our own seed corn. The harvest looks fine this season, but the question is what we plant next year and in the future.

The Experience Paradox

There is a phrase I have heard from hiring managers that used to be a cliché and has become a crisis: "We need someone with three to five years of experience for this entry-level role." It was a joke once. It is now simply the job description.

The logical absurdity is plain. You cannot have a workforce of experienced technologists if you do not create the conditions under which technologists gain experience. Experience is not a credential that can be purchased or a module that can be installed. It accumulates through repetition, failure, feedback, and time. It requires, at some point, being the least experienced person in the room and being allowed to learn from that position.

The AI-assisted senior technologist is not a long-term solution. It is a bridge that leads nowhere if we do not build the other side.

What I Would Change

This is not an argument against AI in the workplace. AI tools are genuinely powerful, and their responsible adoption can make skilled workers more effective. The problem is not the technology. The problem is the business decision to treat AI as a headcount replacement rather than a force multiplier.

The organizations that will fare best in the long run are those that use AI to expand what their teams can accomplish and not to shrink the teams themselves. That means continuing to hire junior talent, pairing them with AI tools that accelerate their learning curve, and investing in the mentorship structures that allow expertise to transfer from one generation to the next.

It means treating entry-level roles not as a cost to be minimized, but as an investment in institutional continuity.

Policy has a role here too. Workforce development programs, apprenticeship incentives, and university partnerships need to be updated for an AI-augmented economy, not designed around an economy that no longer exists, but also not designed around an AI utopia that doesn't exist either. The gap between those two fictions is where the generation of workers is currently stranded.  The bridge must be rebuilt to withstand the innovation that AI has brought to the industry so that the next generation has a path to their future in the technology industry.

The Students Deserve an Answer

When I stood at the front of that classroom at WSU, I told the students the truth as I understood it: the path is harder than it should be, and that is not their fault. The industry made decisions without thinking through the consequences, as industries often do. But the consequences are real and they fall disproportionately on the people who were just arriving.

The students in that room were not asking for a guarantee. They were asking for a fair shot. A chance to do the work, make the mistakes, build the knowledge, and become the experts that the next decade will urgently need.  AI cannot give them that. Only we can and by choosing to build industries that still make room for people who are learning.  If we don't, we will have optimized ourselves into a corner: a small cohort of overextended experts, supported by AI systems that are slowly losing the human knowledge they were built on, with no pipeline to replenish either.

I have joked for years that I will be brought back from the dead in the year 3020 and the first question I'll be asked is, "I hear you know Oracle?" It's a joke born from watching a specialized skill set become so rare that the people who hold it become mythological. There are so few with deep Oracle expertise today that the punchline lands every time I tell it.  The more AI takes hold, the less it feels like a joke. Because what was once a quirk of one monolithic database platform is becoming the trajectory of the entire industry. We are concentrating expertise at the top, eliminating the conditions under which new expertise is built, and trusting AI to fill the gap, all without asking what happens when the AI has nothing left to learn from.

That is not a future I want to help build. And frankly, it is not a future any of us should want to ship, unless we are comfortable with the idea that the most valuable skill a technologist can have in 3020 is simply having been alive long enough to remember how any of this actually worked.

~Peace Out