Who Trains the Next Senior Engineer?

Every company wants to hire someone who already knows what they are doing. Give them the problem, let them figure it out, and get productive work that moves the needle without consuming too much of everyone else’s time. That’s the dream. Hiring someone who has to learn on the job, not so much.

But if everyone wants the person after they have acquired that ‘senior-level’ experience, who gives them the work to acquire it?

Senior engineers have to come from somewhere.

Engineers have to be given assignments, make mistakes, and receive feedback, then repeat that cycle and gradually take on more responsibility for less-defined problems. Some engineers progress quickly because they reason intuitively. Others can still reach that ‘senior’ status and be valuable contributors, but need a little more coaching and exposure before things start to make sense. Either way, being a senior engineer almost by definition requires an opportunity to participate in work whose consequences they can observe, not just knock out tickets with no further context.

Historically, a company could still get useful productivity while that development happened. A junior engineer completed bounded tasks, an experienced colleague supplied direction and review (and hopefully some context and coaching), and the junior gradually learned more about the system. The arrangement produced work today and capability for tomorrow. It wasn’t usually called an apprenticeship, but essentially served that function.

Now consider what happens in environments where the implementation gets substantially faster. Whether through AI-assisted work or simple process improvements, the organization wants to produce more with the people it already has. But someone still has to evaluate the situation, decide what to produce, and assess whether the result delivers a truly effective business outcome (not just a technical delivery). In these cases, insight and judgment become the bottleneck. (I discussed this in more detail in “The Cost of Output Is the Judgment It Consumes.”

Under those conditions, an obvious solution is to shift the load toward people who can operate with greater independence. Staff up on senior engineers and reduce less-experienced, junior engineers. An experienced engineer may cost more, but they can assess the situation, decompose the work, and move to a solution without requiring someone else to make every consequential decision. They cost more, but they free up even more expensive senior attention elsewhere in the organization.

That creates a potential squeeze for juniors. Especially if they are AI-powered, the work they can perform becomes easier to produce, while the frequency and level of guidance they need becomes relatively more expensive. Their salary might be lower, but the full cost includes someone experienced spending time framing, checking, and correcting. 

Those intuitive, reasoning, high-potential juniors can still be compelling in this environment. They practically mask their lack of experience while showing the ability to reason toward an unfamiliar model, make lateral connections, and recognize where their understanding is incomplete. But to demonstrate that potential, even exceptional engineers need a little more context and time with the senior engineer, rather than being heads-down working through tickets. Senior engineer time that is already an expensive bottleneck.

In almost all skilled roles, someone with less experience needs more exposure before they gain that insight and can be trusted to make judgment calls reliably. Some may never develop the capability the role requires, but many can develop it if given the right work and guidance. The employer has to make at least a minimal investment before knowing what outcome it will get, and improving the process for identifying engineers with potential is one of the things that separates great companies from also-rans.

But, especially when an AI agent can act in the place of a junior engineer, the temptation to move to a primarily/exclusively senior-level staff arrangement to increase output and efficiency is very real. However, this brings with it a critical challenge. If companies become less willing to accept the uncertainty of giving a task to someone who might need coaching, people can lose the opportunity to develop before anyone knows how far they could have progressed. This robs companies of future senior-level team members they could rely on but never take the time to develop.

At the level of an individual firm, hiring staff that is already senior can be a perfectly sensible decision. At the market level, it raises a supply question. If everyone is choosing to bypass junior engineers in favor of more empowered senior talent (AI-backed or not), it may work for a time, but eventually those experienced engineers retire, move into other roles, or leave the work entirely. If demand for their capability remains strong or grows, while fewer replacements are developing, that scarcity will affect compensation. At the extreme, even compensation can’t solve a lack of a talent pipeline, and the overall amount of work an entire sector can do could be hamstrung. 

At some point, the calculation may change. Taking the long view and paying for an apprentice’s development may become more attractive than competing for a limited supply of experienced staff. The company accepts lower immediate contribution in exchange for the possibility of building capability it will otherwise have to buy at a higher price. A more explicit apprenticeship model can emerge because the economics begin to support it.

But once this tendency to cut investment in junior staff takes hold, a compounding challenge emerges: rising senior compensation also makes teaching time more expensive. This will drive companies further into the cycle of keeping the mentor focused on solving an immediate business problem. And even for firms developing junior staff, in a senior-premium environment, once they gain experience, the apprentice may leave before the company can recapture much of their investment. Developing people has a delayed and uncertain return. The organization needs a credible way to turn supervision into increasing independence and a reason to believe it will benefit from that progression.

This scenario increases the importance of the apprenticeship’s intentionality and design. Repeating simple tasks does not automatically teach someone how to frame an ambiguous problem. They need to see why a problem was categorized a certain way, which mental model was selected, and how the work was decomposed. They need opportunities to propose their own reasoning and receive feedback before an incorrect approach becomes a large implementation. Companies need to efficiently identify engineers with the potential to reach more senior levels and ramp them up quickly.

Faster tools, like AI, can help. An apprentice can explore alternatives, test ideas, and encounter consequences sooner. But completing more exercises or generating more code is not sufficient evidence that judgment is developing. The question is whether they increasingly understand what they are doing, why it fits the problem, and where it might fail. It will take an experienced human – perhaps aided by their own tools – to evaluate the progress of the individual. The apprenticeship succeeds when responsibility can expand without supervision expanding at the same rate.

None of this implies that junior opportunities must permanently shrink. Lower development costs can create demand for work that previously made little economic sense. Better tools can help people learn, and better training can reduce the cost of developing them. These forces can alter the balance. The argument is that an initial shift toward existing senior capability creates pressures that the market must eventually address, rather than a fixed outcome for every engineer.

There may be a period when companies strongly prefer people who already possess judgment and are less willing to fund its development. Then there may be a period when scarcity, departures, and higher compensation make apprenticeship attractive again. The adjustment takes time because the capability takes time to build. A salary increase can change an incentive quickly; it cannot instantly give someone years of useful exposure.

Companies can hire the senior people they need today. But that level of talent will remain available only if someone gives less-experienced engineers the opportunity to become them. If judgment is becoming more valuable, the process that produces it has economic value too. Eventually, the market has to reflect that value in what it’s willing to spend on development.