The Roles AI Lets You Cut Might Be the Ones Building Your Bench

This week's edition was written by Graham Thornton, President of Consulting and Growth at Talivity.

There is a quiet trade going on inside a lot of TA functions right now. Automate the entry-level work, show the board some ROI on AI, book the savings this quarter. It is clean on a spreadsheet. The bill comes years later, in a leadership bench nobody noticed they stopped building.

I should be the last person raising this. For about a year I have been arguing something close to the opposite: that most of what we call recruiting work was never real work to begin with, a pile of loosely related tasks that exist because someone has to do them, not because the job needs to exist. Point a capable model at that pile and the hollowness shows. AI did not invent that problem, it just made it visible. I still believe it. But a piece in HBR this week caught on something I had been skating past.

Jenny Fernandez opens with a CHRO who cut a 200-person entry-level analyst program to show ROI on AI. The savings landed right away. Eighteen months later the bench was empty. Senior managers were absorbing work nobody had been trained to take over, and the next director cohort had nowhere to come from. She described it as solving a cost problem and creating a much bigger one in its place.

Here is what stuck with me. Those entry-level roles probably looked like exactly the kind of low-value, automatable task pile I have been telling everyone to question. Pulling data. Formatting the deck. Running the first cut of an analysis a model now does in seconds. They fail an efficiency audit. The argument I have been making would tell you to cut them.

But output was never what those roles were for. They were where people learned to sit in the room where decisions got made, navigate ambiguity, and build the judgment no training module installs. The output was the byproduct. The development was the point. Automate the output and you can delete the apprenticeship without noticing, because the apprenticeship never showed up as a line item.

See what AI says about your company.

Our AI search optimization practice opens with a diagnostic that maps where your employer brand shows up across ChatGPT, Claude, and Gemini, where it goes missing, and where third parties are answering for you. See how it works.

The World Economic Forum's January briefing on entry-level work, built on a survey of more than 9,000 early-career workers across 48 economies, lands in the same place from the worker's side. The routine tasks getting automated first are the ones that used to give newcomers their foothold, where they learned to read a room and build judgment before anyone trusted them with the real thing. Its own panel of experts put the fix plainly: automate the repetitive work, protect the learning environment.

That distinction matters twice over for anyone running talent.

It applies to your own team first. The junior sourcer, the coordinator, the recruiter working reqs you are eyeing for automation are also your pipeline for senior recruiters. Cut the rung and you have hit this year's efficiency target by borrowing against your bench five years out.

It applies just as much to the market you hire from. Korn Ferry found 43% of companies plan to replace roles with AI. If enough employers thin their entry-level ranks at the same time, the external supply of mid-level talent thins right behind them. The capability debt is not one company's balance sheet. It is the industry's.

So I am sharpening the argument. AI is still exposing work that should not exist, and most TA functions carry more of it than they would admit. But “a model can do this task” and “this role should go away” are different sentences. The companies that succeed in the coming years will be the ones who can tell those two apart, automating the busywork while protecting the rung underneath it.

Where exactly that line sits will look different for every team, and it is worth arguing about. But drawing it on purpose beats letting an efficiency spreadsheet draw it for you.

Your Talent Strategy Has to Keep Up with Your AI Transformation – HBR. The piece behind this week's letter. Fernandez's “capability debt” framing is the part that earns its keep: the gap between what your business needs people to do and what your workforce can actually deliver, accruing quietly one automated function at a time.

The Recruitics and Maki Partnership, Broken Down – Hung Lee. Hung Lee of Recruiting Brainfood fame breaks down the new partnership between Recruitics and MakiPeople, aimed squarely at the candidate fraud and pipeline noise swamping recruiters right now. The split is clean: Recruitics fills the pipeline, Maki qualifies it. Worth a watch for the problem alone, which is one every TA team is living.

Has AI Already Killed How-To Nonfiction? – Tim Ferriss. Ferriss pulls his own BookScan numbers and finds his catalog on pace to sell roughly 80% fewer print copies than it did before ChatGPT, then names why: his books were lookup tables, and the chatbot is a better interface to the answers. What he bets survives is the part that was never just information, the sequencing, the voice, the real stories that actually move someone to change. Different industry, same line this week's letter is drawing between the commodity layer AI eats and the human one it does not.

Connect with Author: LinkedIn | grahamt@talivity.com