No corner of the AI and work debate has gone more viral in 2026 than the fate of the graduate job. The doom case has a memorable champion in ServiceNow chief executive Bill McDermott, who has suggested that graduate unemployment could climb into the mid 30 percent range within a few years as AI agents absorb the kind of routine tasks that junior staff once cut their teeth on. It is a striking number, and it has travelled far. The harder question is whether the data supports it.
What the evidence actually shows
The picture is real but more modest than the headline fear. Stanford's Digital Economy Lab found roughly a 13 percent relative decline in employment for workers aged 22 to 25 in the most AI exposed occupations, a meaningful dent concentrated precisely where the alarm is loudest. Big technology firms cut graduate hiring sharply in recent years, and startup graduate recruitment fell too. The entry level rung is genuinely under strain. But a 13 percent relative decline in exposed roles is a long way from a third of all graduates out of work.
The boomerang nobody predicted
The more interesting development is a partial reversal. Fast Company reports that some companies which replaced junior workers with AI are now paying the price and rehiring. The logic that broke is simple. Tasks once handled by junior engineers were quietly absorbed by senior staff on the assumption that AI would cover the gap. AI did speed up shipping code, but designing, testing and dealing with stakeholders still needed people, and there were suddenly fewer of them. Cutting the bottom of the pyramid turned out to overload the middle.
CEOs are hedging
Leadership sentiment has shifted accordingly. Despite the gloomy forecasts, surveys this year found a majority of chief executives, around 67 percent, expect higher entry level headcount in 2026, a sign that the first round of cuts may have been premature. The same executives who automated junior roles are now discovering that an organisation with no juniors has no pipeline, no succession and no one being trained to supervise the agents in the first place.
How to read the noise
The honest summary is that AI is reshaping the bottom of the labour market rather than deleting it. The roles most exposed to automation are contracting, the experience gap between education and employment is widening, and the firms that cut fastest are learning that human judgement is harder to replace than human throughput. For this year's graduates, the advice that keeps surviving contact with the data is to learn to direct AI agents rather than compete with them, because the employers doing the rehiring are looking for exactly that.
