If AI writes the first draft, who gets to make the first-draft mistakes that teach a junior what good looks like?

The central question is how junior professionals become senior when some of the work they once learned from is automated.

A large amount of early-career work has historically been repetitive for a reason.

Junior designers produce variants. Junior developers fix small bugs. Junior writers create first drafts. Junior analysts clean data and prepare summaries.

The work is not always glamorous, but it creates exposure to real systems.

AI tools can assist with or generate portions of that first-draft work. How much they replace in a real team depends on the task and the review process.

That creates a productivity gain and a training problem.

If a senior designer can generate twenty layout options without involving a junior, where does the junior learn why nineteen of them are wrong?

If an AI coding agent handles straightforward tickets, how does a new developer build the debugging instincts that later allow them to diagnose a difficult production issue?

The answer cannot be to preserve inefficient work purely as an apprenticeship ritual.

But organisations need a replacement learning model.

Junior roles may need to become more observational, evaluative and systems-focused earlier.

Instead of only producing output, a junior might be expected to review generated alternatives, explain trade-offs, run tests, trace errors, document decisions and pair with senior staff on higher-level work.

That may raise the entry bar if employers expect judgment before they have provided a way to learn it.

It may also make good mentorship more important.

Jobs and Skills Australia’s Gen AI study emphasises adaptation and skills rather than simple displacement. My inference is that employers should plan explicitly for junior training rather than assuming experience will accumulate by itself.

If AI removes the lowest-complexity tasks, companies cannot assume experience will accumulate automatically.

They will need to design for it.

Agencies face the same issue.

The industry has often relied on juniors doing production-heavy work while senior staff manage clients and strategy.

AI compresses that production layer.

A better model may involve smaller teams where junior staff work closer to strategy and systems from the beginning, with AI used as a supervised accelerator.

That could produce stronger professionals.

But only if organisations invest in teaching judgment rather than just measuring output.