“Stop training people and you run out of seniors”

Deloitte is shrinking fast. The striking part: the number of staff under thirty is dropping quickly, and the firm's chief executive acknowledges AI plays a part in that. On the evening of Friday 19 September, Cas Bogaard joined NOS Met het Oog op Morgen on NPO Radio 1 to explain what is going on. Presenter Elisabeth Steinz pushed on the question underneath it: if AI takes over the work beginners do, where do the experienced people come from?

Why junior roles go first

Most people still use ChatGPT as a kind of Google. They look something up, or draft a paragraph with it. That picture is out of date. An AI model can now work for hours on end: analysing, writing, coding.

“That wipes out a lot of the tasks where juniors are normally allowed to make mistakes.”

Cas Bogaard, NOS Met het Oog op Morgen, September 19, 2026 (translated from Dutch)

That is exactly the kind of work at the bottom of the ladder. Executional, repeatable, checkable. The work you learn your trade on. In professional services there is a second factor: the billable-hour model is under pressure. Once you sell outcomes instead of hours, a large team of juniors turns from a revenue engine into a cost line.

This is much wider than Deloitte

The trend runs across the whole economy, and hardest of all in banking.

“In banking in particular, an enormous number of jobs are disappearing. ABN AMRO, 25 percent of jobs gone, AI a major factor. Same at ING, at Triodos, but really right across the board.”

Cas Bogaard

The logic is not complicated. It is cheaper and it is more efficient. And when shareholders apply the pressure, that is a hard argument to push back on.

The sum that does not add up

More efficient is not automatically bad. You could just as easily say: if AI takes over that work, those young people end up somewhere else, possibly somewhere more interesting. That is the expectation longer term. In practice, though, hardly anyone is looking at it that way yet.

Meanwhile something else is heading towards these organisations. The EU AI Act takes full effect at the end of December 2027. One of its biggest components: a human has to stay in the loop on AI workflows. Human oversight is not a nice-to-have, it is a requirement.

“Who can actually do that oversight work well? You need experience for it, you need instinct for it, you need to have made mistakes.”

Cas Bogaard

That calls for mediors and seniors. And there is the knot. Mediors and seniors are juniors who grew into the role. Cut the junior layer and in five to ten years you have also cut the layer that is supposed to do the checking.

“How do you end up with people who have more experience, if as an organisation you are not training anyone at all?”

Cas Bogaard

Why companies do it anyway

Companies are not stupid. They know training matters, and some genuinely invest in it. But it varies a lot from firm to firm.

“A great many companies simply cannot escape shareholder pressure. So they pick the short term, the profit, over investing in your people long term. Which is how you build a durable business.”

Cas Bogaard

A junior quota, or a subsidy

In Het Financieele Dagblad and de Volkskrant, Cas argued earlier for a legal junior quota: a fixed share of junior staff a company has to employ. On air he qualified that.

“It is mainly a provocative idea. Companies are under pressure because they have to deliver profit. They lack the incentive to think strategically about the longer term.”

Cas Bogaard

The quota is one possible instrument. A subsidy on junior positions could work just as well. The point is not the instrument. The point is that something has to counterbalance the short-term incentive.

HR belongs at the table

The fix that sits inside organisations themselves is the less obvious one.

“It sounds a bit counterintuitive, but HR deserves a seat in that AI discussion.”

Cas Bogaard

In most AI transitions, HR is not in the room. The conversation is about tooling, about data, about compliance. Yet the questions that matter are workforce questions: how do we treat our people, which skills do we need tomorrow, what does the organisation look like in three years. That is the step a lot of organisations skip.

That is exactly where the work of CribConnects sits. Not just building, but thinking through how you absorb the AI shift as an organisation. Responsible adoption, with a plan for the people who have to make it work.

The full conversation is available on NPO Radio 1, or as part of the full episode on Spotify. The broadcast is in Dutch.