Your Staff Won't Lose Their Jobs to AI. They'll Lose Their Pay Rises Instead.

Everyone tracking Malaysia's 697,000 jobs at risk from AI is watching the wrong number. The real effect isn't showing up in the headcount. It's showing up in the paycheque.

On July 30, Apollo Global Management economists published a study covering 321 occupations across the US labor market. Their finding: employment in high-AI-exposure jobs stayed stable. Real wages in those jobs fell 6.7% since 2023 — the year ChatGPT went mainstream. Among service workers in the most AI-exposed roles, the decline hit 24.3%. Across 5.8 million workers in those jobs, wage compression amounts to approximately $28 billion in lost annual earnings. Not lost jobs. Lost raises.

The Malaysian context runs parallel. HR minister Sivakumar reported 42,807 actual job losses in H1 2026. The 697,000 figure gets the headlines. But the slower story — workers who kept their jobs absorbing higher productivity expectations without equivalent pay growth — isn't being tracked with the same precision.

Who this really matters to:

→ Malaysian employees in high-AI-exposure roles — customer service, administrative work, content writing, data entry — the Apollo finding isn't that you'll be replaced; it's that your market rate may quietly decline as employers learn what an AI-assisted person can produce → Malaysian HR directors running 2027 salary reviews — if your team's output increased this year because of AI tools you gave them, the productivity benchmark for the next pay discussion has shifted; that's both a business advantage and a retention risk → Malaysian fresh graduates entering AI-exposed fields — starting salary benchmarks may already be under pressure; the offer you negotiate today reflects a productivity floor your employer's AI tools helped reset → Malaysian business owners who rolled out AI tools this year — if your team delivers more output per person, where that gain goes is a management decision you're already making, even if you haven't framed it that way

MULTIPLE PERSPECTIVES

The Apollo mechanism is specific: AI-assisted workers produce more, so employers can maintain output with fewer hires, or demand more output without raising wages. In practice, most companies aren't running layoffs. They're managing attrition, filling fewer vacancies, and capturing the productivity gap in margins. The worker who stayed kept their job. The invisible effect is the one who didn't get hired at all — who never shows up in job-loss statistics.

For Malaysian employers, this is a choice point, not a market outcome. Your team's effective productivity improved because of AI tools you deployed. Whether that improvement flows into margins, growth, or better compensation is a decision you're making now.

The uncomfortable implication: if AI productivity gains consistently flow to employers rather than workers, and Malaysian employees in affected roles begin to notice — the retention dynamic changes before any layoff data does.

If AI made your team 30% more productive this year — who captured that 30%?

If it went into higher output with the same headcount and pay — your team already knows; the retention risk doesn't appear in your numbers until it does.

If you shared the gain through better pay, reduced hours, or lighter workload — you're a minority in the Apollo data, and that difference is becoming a visible hiring and retention differentiator.

The 697,000 number counts jobs at risk. The question worth asking is what those jobs are paying the people who keep them.