Global Talent #57

Brought to you by Lundi: we design, hire, and run international teams.

🔥 Opening Shot

Every good team I have ever built had a junior engine inside it. Kyiv, Warsaw, Guangzhou, it never mattered where. Somewhere in the org there were twenty-two-year-olds doing the unglamorous work, getting reps, breaking things, and turning into the people who ran the place five years later.

Stanford just put a number on what is happening to that engine in America. Among workers aged 22 to 25 in the jobs most exposed to AI, employment now sits about 19 percent below where it would be if it had kept pace with their less-exposed peers. Not because companies are firing juniors. Because they have quietly stopped hiring them.

I understand the spreadsheet logic. An AI subscription costs less than a graduate, and the model already knows everything a textbook knows. Why pay someone to learn?

Because the textbook knowledge was never the point. Seniors are made of reps: the client call that went sideways, the migration that failed at 2 a.m., a thousand small judgment calls no model has seen. Cut the first rung and you have not automated the ladder. You have scheduled a senior shortage for 2031, and you will pay for it in a bidding war.

The companies that get this will keep building benches. They will just build them where training a junior still pays. The entry level is not disappearing. It is relocating. Your future senior team is now a design decision, not a market condition.

This Week's Number: 19% — how far employment of 22-to-25-year-olds in the most AI-exposed US occupations has fallen behind their less-exposed peers since ChatGPT launched.

📌 On the Radar

1. Stanford: no AI jobs apocalypse, but the first rung is cracking.
The Stanford Digital Economy Lab published an updated version of its "Canaries in the Coal Mine" study on August 12, tracking millions of ADP payroll records through June 2026. Economy-wide, the authors find no widespread AI job displacement. But for workers aged 22 to 25 in highly exposed occupations, the employment gap versus less-exposed peers has widened from 15% a year ago to 19% now, and it operates through reduced hiring, not layoffs. The mechanism is the sharpest part: employment is falling in jobs built on codified knowledge (textbooks, documentation) while rising for experienced workers in jobs built on tacit knowledge earned through practice.

AI is not eating jobs. It is eating the on-ramp to jobs, and it is doing it first in the markets where a graduate costs the most. Skipping the junior is individually rational and collectively a pipeline failure, because every senior you will hire in 2031 has to be a junior somewhere in 2026. The operators who win this will run the numbers the other way: pair the AI with supervised juniors in a market where training still pays, under experienced leads, inside a properly employed and managed structure. Starting that build with an employment partner is the right first move; turning it into a senior-production line is operating work, and that is the half most companies leave to chance. The cheapest senior you will ever hire is the junior you develop now, somewhere the math still works.

2. The jobs beneath your plan were revised away. Again.
On August 28 the Bureau of Labor Statistics published its preliminary benchmark revision: total US nonfarm employment through March 2026 was marked down by 79,000, and private payrolls by 178,000. The benchmark reconciles the monthly estimates against unemployment-insurance tax records that nearly every employer must file, the closest thing to a full count that exists. Forecasters had expected an upward correction this year. They got another markdown.

Every operating plan quietly assumes the labor market described in the headlines actually exists. Lately the errors run one direction. If your hiring model, comp benchmarks, and backfill assumptions are all calibrated to one country's payroll data, your plan inherits that country's measurement error on top of its market risk. Concentration in a single talent market is also a data risk, and the fix is the old one: build across more than one market, and plan from numbers you can verify on your own payroll.

3. Poland is about to make wage planning boring again. Enjoy it.
Poland's government has proposed a 2027 minimum wage of PLN 4,950 per month, a 3.0% increase, with draft-regulation groundwork filed on August 25 and a final decision due by September 15 after talks with unions (who wanted at least PLN 5,200) ended without agreement. For context: the floor jumped 21.5% into 2024 and 10% into 2025, then just 3% into 2026. A second straight ~3% year would end Poland's wage-panic era for good.

For a white-collar build in Kraków or Wrocław the statutory floor itself rarely binds. The signal is what matters. Poland spent three years as Europe's poster child for runaway labor-cost inflation, and plenty of CFOs crossed it off the list on that trend line alone. The trend line just went flat, and the talent pool is as deep as ever. Note the fine print, though: several statutory cost items key off the minimum wage, and Poland's labor inspectorate gained real contractor-reclassification powers in July. Costs are stabilizing exactly as structural scrutiny rises, which is precisely when the properly built team regains its edge over the improvised one.

📊 Chart of the Week

One cohort, two different labor markets. Since ChatGPT launched, employment of 22-to-25-year-olds has fallen about 11% in the most AI-exposed occupations while rising about 10% in the least exposed, and the adjustment runs through hiring freezes, not firings. The authors call these patterns descriptive, not proof of AI causation. An operator does not need causation settled: the domestic pipeline of trainable juniors in exposed functions is thinning, and whoever still develops that talent, wherever it pays to do so, owns the senior market in five years.

🚀 One More Thing

Here is a question worth an hour this quarter: where does your next generation of senior people actually come from? If the honest answer is "we will hire them from the market when we need them," you are betting your 2031 org chart on a market everyone else is quietly exiting. If you want to think through what a deliberate bench looks like (roles, market, structure, cost), grab a strategy session and I will walk you through how we design junior-to-senior pipelines for the teams we run. No pitch. Just insight from someone who has been there.

📖 New here? My book, Winning the Global Talent War, is the full playbook behind this newsletter.