AI May Have Broken the Apprenticeship Model That Built Software
For decades, hiring junior developers was one of the safest investments a software company could make. They cost less, handled lower-complexity work, and gradually developed into the next generation of senior engineers. The mentoring investment was real, but so was the payoff.
I'm no longer convinced that's true.
One of the more interesting studies I've seen came from Jellyfish, which found senior developers using GitHub Copilot completed coding tasks roughly 22% faster while junior developers saw gains closer to 4%. Fastly reported a similar trend, with senior engineers producing significantly more AI-generated code than their junior counterparts. The pattern is becoming fairly consistent: AI appears to amplify experience more than it replaces it.
That's where things get difficult for smaller software shops.
Historically, junior developers generated value while they learned. They handled lower-complexity work, absorbed knowledge from senior engineers, and gradually took on more responsibility. Today, much of that lower-complexity work is exactly the type of work AI is getting good at. The mentoring cost remains. The productive output is shrinking.
At the same time, workforce incentives have changed. Deloitte's Gen Z and Millennial Survey continues to show younger workers prioritizing career growth, advancement opportunities, and mobility. The expectation that someone will stay with a company long enough for that training investment to fully pay off is becoming harder to justify.
Put those trends together and the math starts to look very different than it did ten years ago.
A 20-person software shop hires a junior engineer. Their senior developers absorb the mentoring burden. The team sacrifices productivity while that engineer develops. Then, just as the investment starts paying dividends, a larger company hires them away with a title change and a meaningful raise.
Nobody is behaving irrationally.
The developer is maximizing their career. The larger company is buying proven talent. The small shop is the one expected to absorb the cost.
That's where this conversation usually turns into a moral argument. Someone needs to train the next generation. Someone needs to create future senior engineers.
Maybe.
But small software companies aren't universities. They're not workforce development programs. They're businesses trying to survive, grow, and compete.
The uncomfortable reality is that AI is increasing the value of experienced engineers at the exact same moment that workforce behavior is reducing the return on developing junior talent. One force is making senior engineers more productive than ever. The other is making it harder than ever to justify the investment required to create them.
Everyone agrees we need future senior engineers. The developer wants career growth. The large company wants proven talent. The small company wants productivity. All three are behaving rationally.
The problem is that only one of them is being asked to pay for the pipeline.
Everyone agrees we need future senior engineers. The question is who is going to pay to create them.
Right now, the answer seems to be someone else.
Sources:
• Jellyfish – AI Codegen Tools Propel Senior Developers
• Deloitte 2025 Gen Z and Millennial Survey
#AI #SoftwareEngineering #EngineeringLeadership #DeveloperExperience #TechLeadership #FutureOfWork #ArtificialIntelligence