Mauritius complains often about brain drain, about those graduated Mauritians who find employment outside of Mauritius. However, last week, I read a paper from HEC Paris and the Banque de France that I feel every HR director in Mauritius should read. It follows almost every permanent contract signed and ended in France’s private sector, month by month, since 2019.

The finding is not the one the headlines prepared us for. Since ChatGPT arrived in November 2022, French firms have not been firing young people in the jobs most exposed to AI; they have been hiring fewer of them. Permanent hires of 21- to 29-year-olds in those jobs are about 28% below their late-2022 level, against 8% in the least exposed jobs. Departures fell too. So, the authors call it a freeze: fewer people in, fewer people out.

The firms that say they use AI intensively show the same pattern. They recruit relatively fewer young people on permanent contracts, while their total headcount moves no differently from everyone else’s. Nothing shows up in the annual report. The change happens at the door.

I went looking for the Mauritian equivalent and could not find one. Youth unemployment stood at about 17% in 2025, roughly three times the national rate, according to the World Bank’s estimate. I found no public series that tells us how many graduates our banks, management companies and back offices did not hire this year.

Read the French list of the most exposed jobs with Mauritius in mind: banking operations technicians, financial and economic analysts, translators, document processors. Many of the jobs that bring our graduates into global business, banking, outsourcing and accounting look like that list.

The cause is contested. The consequence is not. Economists do not agree on why junior hiring is slowing. A Stanford team, working from US payroll data, sees employment of 22- to 25-year-olds in AIexposed jobs shrinking by almost 4% a year, and says the effect survives every alternative explanation it tested. A Warwick team, working from 650 million hiring records in four countries, says the real culprit is remote work, and that AI’s effect largely disappears once you account for it. Norway’s complete employment records show almost nothing yet.

For an HR director, the argument matters less than it seems. Whether the cause is an AI model, a hybrid policy or interest rates, a company that hires fewer juniors trains fewer of them. Whatever the cause, a company that stops hiring juniors today is deciding who will not be its manager in 2032.

Three European economists formalized the logic this summer. When automating junior work saves only a little money, it lowers output in the long run, because fewer people climb the ladder that produces experienced staff. Today’s seniors gain, tomorrow’s seniors never arrive.

Five questions to put to your team this quarter.

How many juniors did we NOT hire? Ask for first job hires per year since 2021, by department, and the ratio of juniors to seniors. Put that figure in front of the board once a year, next to total headcount. A flat headcount can hide an empty bench.

What did the junior learn from the task we automated? If AI now drafts the first version of the credit memo, the reconciliation note or the client report, ask who still learns to write one. Name the person who can do it by hand today, and the person who will be able to in five years.

Did we rewrite the entry job or delete it? PwC’s 2026 analysis of job postings finds that entry roles in AI-exposed jobs are seven times more likely to ask for skills that used to come later in a career: judgment, stakeholder management, decision-making. If you raise the bar, build the ramp. Reviewing AI output alongside a senior is how a junior learns judgment, and it should be budgeted as training.

Is our hybrid policy doing the damage we blame on AI? The Warwick finding is a warning about onboarding. A graduate who spends the first six months alone with a laptop learns the tool and misses the trade. Ask whether every new hire spends their first months in the same room as the people they learn from.

Are we using what already exists? Mauritius has run graduate placement schemes for more than a decade, and the Bank of Mauritius runs a two-year Young Graduate Programme that lists data analytics and AI among its fields. A structured programme costs less than a succession plan with no names in it.

A disclosure: I sell AI services, so I have an interest in companies adopting AI. That is exactly why I want them to adopt it without burning their bench.

The French data show how quiet this is. Nobody protests over a job they never got. No union negotiates for the graduate who was not called back. The cost shows up in ten years, when the people who should be running your operations spent their twenties somewhere else.

Count your juniors this year. Nobody else will.