PwC’s AI assurance leadership has floated something that sounds almost backwards: new hires stepping into reviewer and supervisor roles almost immediately. That raises an obvious question — reviewers of what, exactly, if they’ve never done the underlying work themselves?
The instinct is to say we need a whole new training program. Here’s the case for why that’s the wrong instinct — and what to do instead.
Why the Old Pipeline Doesn’t Work Anymore
The old apprenticeship model ran on a byproduct: a reviewer marks up a junior’s draft, and in explaining why something was wrong, they’re forced to externalize judgment they’d otherwise never write down. That correction existed for one moment, taught one person, and vanished. People were learning by doing, not by studying — the classic 70-20-10 split, where roughly 70% of learning comes from just doing the job.
That system assumed years of rote, repetitive work — vouching invoices, tie-outs, reconciliations — before judgment really kicked in. The rote work was never really about the invoices. It was the training ground where pattern recognition got built, almost as a side effect of doing something boring enough times to start noticing when something was off.
AI now does most of that rote work faster than any junior could. And Stanford and MIT research found something worth sitting with: junior accountants tend to accept AI output with comparatively little scrutiny, while experienced staff treat AI as a collaborator to be actively challenged. Those years of rote work were quietly building the instinct to ask, “wait, does this look right?” Compress the years away, and the instinct doesn’t automatically come with it. The people with the least trained judgment right now are the ones least likely to question the system handing them answers.
Why You Can’t Wait for Schools to Fix It
Schools are responding. AACSB, one of the top three global business school accreditors, added a Digital Agility standard requiring programs to teach students to interpret AI output with accuracy and skepticism. That’s real, and it’s good.
But it’s a multi-year curriculum change working its way through university programs. Your next hiring class is arriving now — long before any of that filters through to a diploma. You can’t outsource this timeline to higher ed. Whatever gets built has to get built on your side, on the firm’s clock.
The pressure backing this up: BambooHR data shows firms hiring at roughly a 3-to-1 senior-to-entry ratio, meaning far fewer entry-level roles exist relative to experienced hires than before. Each entry-level hire gets pushed into higher-responsibility work faster, with less runway. A third of new accounting and finance hires quit within their first year, with mismatched expectations cited as the leading reason.
The Fix: Bank What You’re Already Making
Here’s the reframe. Every real correction you make this week while reviewing AI-assisted work — the wrong clause, the missed cutoff date, the reasoning gap — is already a training artifact. The only change needed: don’t let it disappear.
When you’re reviewing AI-assisted work, write down and externalize your thoughts — both when something goes wrong and when it goes right. Redact the client-specific details, and this becomes a checklist. You can even use AI to take those notes and turn them into a structured doc. Build one file per process or engagement type.
Over a few months, that file becomes a real, growing, accumulated set of actual mistakes and good practices in your specific area — caught by your own people, explained by your own reviewers. A new hire can use it to find and explain errors themselves, the same way reviewers used to test judgment before AI existed. They don’t have to have done the work before; they have a long list of the good and the bad for that one thing.
This is exactly how firms will differentiate themselves from an AI perspective. If everyone is using the same underlying models, the secret sauce becomes the context you’re able to give it. A list of where AI has done a specific task poorly or well in the past becomes proprietary context — it improves your outputs, and it tells a new hire exactly what to look for from day one.
Key Takeaways
- New hires are being pushed toward reviewer-adjacent roles faster than the old apprenticeship pipeline ever assumed.
- Junior staff scrutinize AI output less than experienced staff do — exactly backwards from what a compressed pipeline needs.
- Schools are responding, but the timeline won’t help the hiring class in front of you right now.
- The fix: bank the real corrections you already make during review — anonymized, in one file per process — instead of designing training from scratch.
Want the CPE credit? Take the full lesson on EverydayCPE and earn 0.2 CPE credits: https://everydaycpe.com/courses/non-technical-personnel-human-resources/the-curriculum-youre-already-writing/
Leave a Reply