This deep dive expands on the Loop Asia conversation with Keith Carter, Founding Partner at KDA Capabilities.

Jon mentions it almost in passing — some companies have "stopped hiring new graduates for a while" — and Keith Carter doesn't let it pass. "That's right and yet though you know that's such a short-sighted view," he says. "Really you're gonna have the AI run do everything and not yet. It can't." Underneath those two sentences is a workforce-planning mistake a lot of COOs and Heads of Professional Services are making right now.

The freeze looks defensible on a spreadsheet: graduate hires are expensive to onboard, slow to become productive, and AI can already do a meaningful share of what a first-year hire used to do. Carter's objection is structural: AI "can't" yet run the work without oversight, and the person who checks the output and catches the edge case still has to exist. A freeze doesn't remove that need — it removes the pipeline that produces the people capable of filling it.

The horseshoe maker's real lesson

Carter's own analogy, built for individual careers rather than hiring policy, makes the mechanism clear. A horseshoe maker who didn't become a tire changer had no career left — not because the trade vanished, but because "changing a tire takes less skill and less effort than making and changing a horseshoe," so more people could do it, and the economics of the old skill collapsed. Read that organizationally: the winners retrain fastest into the adjacent, still-scarce skill — Carter's examples are "programmer to AI wrangler" and a copy editor moving from single documents to hundreds at AI-assisted speed, where the differentiator becomes volume plus quality plus imagination.

That retraining has to be built, and a freeze starves the raw material for it. The graduate who would have spent eighteen months learning a workflow the slow way is exactly who, with the right structure, learns the AI-wrangler version in a fraction of the time. Cutting graduate intake doesn't protect the organization from cost — it protects competitors from having to compete with your bench in three years.

What replaces the freeze

Carter's framework points to redesigning what the entry-level role trains for: which judgment calls still need a human learning curve, which AI-assisted volume work builds that judgment fastest, and how oversight gets deliberately taught rather than assumed to appear on its own. That's a capability-building exercise, not a cost-cutting one — the same distinction Blue Connector makes with clients under an AI mandate: build the client's own team's capacity to make these calls, don't hand them a tool and create dependency. A graduate freeze does the opposite, cutting the input to internal capability exactly when that capability is what's in short supply.