Somebody has found a wonderful way to save on training: hire people after someone else has already trained them. Add AI to the slide, remove the junior roles from the budget and call the missing apprenticeship an efficiency gain.
I can see why the spreadsheet likes it. I am less sure where the next generation of experienced people is supposed to come from.
The October 7 debate about AI and developer careers includes anxiety about smaller teams alongside arguments that computing knowledge still matters. A question circulating on X the next day asks whether business users and agents could eventually remove developers from the process. Both are plausible subjects for discussion. Neither supplies a training plan for an actual organization.
I want to talk about the plan.
What the beginner was meant to learn
There is a weak defense of junior work that amounts to preserving repetitive tasks because repetition existed before. I do not find it persuasive. If a tool can remove a pointless nuisance, removing it may make the day and the learning experience better.
The stronger defense asks what the person was learning through the task. Were they practicing how to read a requirement? Learning where an important value came from? Discovering that two reasonable assumptions can collide in the same system? Getting feedback on how to describe a change to somebody who did not watch it being made?
Those abilities need opportunities, feedback and increasing responsibility. If the old task disappears, the team should decide where the learning moves. Otherwise the organization has changed its staffing model without describing how it will renew its judgment.
An assistant can make this more interesting. A learner can compare possible approaches, inspect a proposed change and ask why it fits the constraints. They can make a small prediction, run a bounded experiment and examine the result. The tool can support the learning loop if the person remains engaged with the question.
The weaker arrangement asks them to produce a plausible result and rewards them for not interrupting the flow. They learn which phrasing gets output, while the team mistakes the output for understanding. That can look productive until a new situation arrives and nobody knows which assumption stopped applying.
The senior job is not merely typing slowly
An experienced colleague often contributes by noticing an ambiguity before it becomes an incident. They remember why a rule exists. They distinguish the case that deserves a quick fix from the one that reveals a wrong model of the problem. Those contributions can be hard to count, especially on a slide devoted to how much code appeared.
That does not mean experience is mystical or that tools cannot help with it. It means the organization should describe the work before declaring it redundant. If the team wants an agent to perform a judgment, name the judgment and explain how the result will be assessed. If the answer is that a senior person will check it, you have preserved the need for the person while weakening the route by which people become that person.
I would make learning part of the delivery plan. Give someone a manageable change to explain. Let them participate in reviewing another approach. Ask which evidence supports their conclusion. Make it acceptable to say they do not understand a generated result yet.
This will not settle the future of the labor market. It can improve the future of the team in front of us. An organization cannot control every economic change, but it can avoid congratulating itself for a skills pipeline it has quietly stopped maintaining.
Use the tools. Improve the apprenticeship. Stop describing the absence of a learning plan as proof that learning is obsolete.
A thought worth keeping?
