AI just automated PR's training ground, and the profession hasn't rebuilt it

I hear a version of this conversation in almost every agency I work with these days. Someone points out how much faster everything moves now – research that used to take a junior half a day gets done by a tool in seconds, media lists that took an afternoon to build by hand now build themselves, a first draft of a press release appears before you've finished your coffee–everyone nods. On paper, this is the epitome of efficiency.

It isn't. It's the quiet removal of the only training ground this profession has ever had.

I've been recruiting PR talent for 30 years, and from the outside, the entry-level job has always looked like admin. Trawling through coverage, building media lists contact by contact, sitting in on client calls with nothing to contribute except your attention, just to hear how your boss handled a tricky question. You'd draft a press release, your senior would mark it up and hand it back, you'd redo it, and this happened over and over. That was the job. 

And it was also, without anyone naming it, the entire curriculum. You learn what a good media list looks like by building fifty average ones first; you learn to tell a pitch that'll land from one that won't by watching a hundred of them get turned down. Slow. Not glamorous. But it did the actual work of teaching juniors the craft.

Now, AI has taken over this exact layer. And because it looks like admin rather than expertise, almost nobody has noticed what's actually disappearing: the way junior talent learns.

There won't be some dramatic moment where the industry wakes up and realises it's out of talent. It'll be slower than that, and harder to spot. Agencies will fill up with people who can prompt a model fluently and have almost no instinct for when the model's answer is wrong – you don't get that instinct from watching AI do the work; you get it from doing the reps yourself, badly, often enough that you start to recognise a mistake before it lands. Take away the reps, and you end up with juniors who never turn into the kind of trusted advisor that's now the one thing AI can't do for a client, which is exactly where the value in this business is heading.

To be clear about what I'm not arguing: I'm not pining for the grunt work. Nobody actually wants to keep manual coverage scans alive for their own sake, and no agency head is going to reintroduce them on purpose just to preserve a training pipeline that used to exist by accident. That sort of response misses the point because the tasks were never the thing that mattered – the learning that happened inside them was. If AI strips the task out, the job now is to build a faster, more deliberate way to teach the judgement, not to defend admin work as if it were sacred.

What that actually looks like deserves more thought than "it'll sort itself out." It probably means putting juniors in front of real judgement calls much earlier than the old system ever did, not because they've earned it after years of drafts, but because that's now the only place left for the learning to happen. It means treating "why is this pitch wrong" as something you deliberately teach in someone's first year, instead of something they absorb sideways over five. It means senior people coaching more, not less, because the coaching that used to happen for free, as a byproduct of shared grunt work, doesn't happen automatically anymore. And it means staying in the room: getting juniors onto client calls, in front of real people, instead of everyone disappearing behind laptops, AI tools, and email threads.

This isn't really an HR problem, even though it keeps getting treated like one. A profession that can't train its own replacements can't scale. It can't hand off client relationships responsibly to whoever comes next. And it loses its best argument for existing at all in an AI-saturated market – that this is a discipline built on judgement, relationships and instinct a model can't replicate. That argument only holds up if there are actually people coming through who have those things. Right now, the pipeline that's supposed to produce them is being dismantled by the exact tool the industry keeps congratulating itself on.

The entry-level rung didn't vanish because someone decided PR no longer needs junior talent. It vanished because the work that used to build that talent got automated before anyone stopped to ask what would replace it. That's fixable, but only if the industry treats it as the structural risk it is, not an efficiency story with no downside.