What Happens To Agency Training Programmes When AI Does The Entry-Level Work?
Most agency training programmes were designed around a simple assumption: new hires learn by doing a large volume of entry-level work under supervision, gradually earning more responsibility. AI has quietly removed much of that entry-level work, and the training built on top of it hasn't caught up yet.
Graduate and trainee schemes at agencies have historically rotated new hires through repetitive tasks, formatting decks, producing basic layouts, drafting first-pass copy, precisely because doing that work many times over builds the muscle memory and instinct that more senior roles rely on. When AI can produce a competent first draft of most of that same work in seconds, simply assigning it to a trainee stops making sense as a learning exercise, because there's no longer a meaningful skill gap between what the trainee produces and what the tool produces unassisted.
The temptation to just skip the step
The easy response is to let AI handle that layer entirely and move trainees straight into reviewing and refining AI output. That is more efficient in the short term, but it skips the part of the old apprenticeship where struggling with a blank page or a difficult layout by hand actually taught something the finished output alone doesn't teach: what makes a choice wrong, learned by making the wrong choice yourself first.
What redesigned programmes are starting to look like
The training schemes adapting well are not eliminating hands-on production entirely, they're compressing the volume of repetitive work while adding structured critique earlier: trainees produce a smaller amount of work manually specifically to build instinct, then spend meaningfully more time than before learning to evaluate, edit and defend creative decisions in front of more senior colleagues. That shifts the centre of gravity of a training programme from volume of output toward quality of judgment, built faster because AI removes some of the grinding repetition that used to eat up the early years.
The old programme taught craft through repetition. The new one has to teach judgment faster, because the repetition that used to build it is now optional.
Why this matters beyond any one cohort
Agencies that get this redesign wrong risk producing a generation of mid-level staff who can evaluate AI output competently but never developed the underlying instinct for why one choice beats another, because they were never made to sit with a bad first attempt of their own long enough to understand what made it bad. That gap tends to surface years later, in senior roles where instinct matters most and there's no longer a training programme around to catch it.
Advertisement
