Margin of Safety #28: AI and Entry-level Jobs

Can anyone accurately predict what AI will do to jobs?

The conversation around AI and jobs continues to swing back and forth. Just a couple of months ago, the fear was that mid-level managers would be next on the chopping block. Now, a new paper (“Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence”) argues something different: that AI is already taking entry-level jobs, particularly among 22–25 year olds in occupations most impacted by AI, such as software engineering and customer service.

Key Findings of the Paper

• “Our first key finding is that we uncover substantial declines in employment for early-career workers (ages 22-25) in occupations most exposed to AI, such as software developers and customer service representatives. In contrast, employment trends for more experienced workers in the same occupations, and workers of all ages in less-exposed occupations such as nursing aides, have remained stable or continued to grow.”
• “Our second key fact is that overall employment continues to grow robustly, but employment growth for young workers in particular has been stagnant since late 2022. In jobs less exposed to AI young workers have experienced comparable employment growth to older workers. In contrast, workers aged 22 to 25 have experienced a 6% decline in employment from late 2022 to July 2025 in the most AI-exposed occupations, compared to a 6-9% increase for older workers”
• “Our third key fact is that not all uses of AI are associated with declines in employment. In particular, entry-level employment has declined in applications of AI that automate work, but not those that most augment it”

Our Takes

So what do we make of this? For starters, the target seems to keep moving. Not long ago, the consensus was that AI would hollow out mid-level management (example #1, #2), and now we’re told those roles may actually expand, while entry-level positions disappear instead. Personally, we don’t believe we know enough to make such bold claims. The limits of AI (or whether AGI is even possible) remain unknown, and predicting downstream industry effects is guesswork. It is impossible to prove that AI will not take jobs, simply because we don’t know. We don’t know how AI will develop, we don’t know how markets will adapt, and we don’t know how humans will respond.

The truth is that there are too many variables in play, many of which are confounding. Is the decline in young workers really due to AI, or could it be a broader macroeconomic chill, or even the shift in interest rate regimes that has reshaped business planning? Like macroeconomic forecasts, the explanations will feel obvious in hindsight, but are almost impossible to disentangle in real time. Software developer demand may be falling, but nearly every company we speak with still wants to hire AI talent. Somewhere in the overlap lies a market for AI-fluent engineers, even junior ones.

As Charlie Munger famous said, “I’d rather be roughly right than precisely wrong”. So, instead of predicting that AI will further decrease jobs of 22-25 year olds in occupations most exposed to AI but increase jobs of 26-50+ year olds (which is oddly very precise), I’d rather argue that AI will bring a sea of change to how various tasks are performed, as we noted on our previous blog post.

Which leads to the natural question: so what?

As investors in AI and security, the uncomfortable reality is that our jobs may be at risk too. Our overall view is that it’s premature to make precise predictions about AI’s impact on the labor market. There are too many moving parts and we frankly cannot forecast the future capacity of AI. The one certainty is that we need to lean in and learn. Ultimately, we are not competing with AI itself, but with humans who are armed with AI.

Reach out to us if you are building in the space of AI & security. We have some thoughts!
Kathryn Shih – kshih@forgepointcap.com
Jimmy Park – jpark@forgepointcap.com

This blog is also published on Margin of Safety, Jimmy and Kathryn’s Substack, as they research the practical sides of security + AI so you don’t have to.