Microsoft's AI chief says all white-collar work will be automated in 18 months. The Bureau of Labor Statistics has news for him: the automation is already here.

This is the first hard government data proving AI job displacement is real and accelerating. Not predictions. Not surveys. Official BLS employment numbers.

18 occupations flagged by the BLS as "exposed to AI" lost jobs for the second straight year. These roles account for roughly 10 million jobs. They fell 0.2% from May 2024 to May 2025 while the broader labor market grew 0.8%.

Strip out medical secretaries (healthcare boom offset) and the other 17 occupations fell 1.6%. For the second year running.

Customer service representatives: down 130,180 jobs (-4.8%) in one year. Secretaries (excluding medical/legal/executive): down 31,030 jobs (-1.8%). Wholesale and manufacturing sales reps: down 28,670 jobs (-2.3%).

Since May 2022 (pre-ChatGPT), credit authorizers are down 26.2%, broadcast announcers down 20.8%, sales engineers down 13.2%. These aren't cyclical dips. These are structural shifts.

Goldman Sachs confirmed the pattern: "Occupations highly exposed to AI substitution have seen job openings fall below pre-pandemic levels." The New York Fed published research showing early labor-market effects in job postings. Harvard Business School found job postings for repetitive/structured occupations decreased 13% since 2022.

Connect this to what we've been tracking. Detroit automakers cut 20,000 jobs. Ford's CEO said "half of white-collar work" is at risk. Challenger reported 49,135 layoffs year-to-date with AI cited as a primary factor.

The BLS cautioned its list isn't exhaustive but represents a "reasonable expectation of AI-driven impact." Translation: this is just the beginning.

So when Microsoft's Mustafa Suleyman claims 18 months until white-collar automation, he's not predicting the future. He's describing the present. The BLS data proves AI isn't coming for jobs. It's already eating them.