The predicted AI jobs apocalypse hasn’t shown up. Something quieter and arguably more insidious has taken its place: your job stays, but your pay shrinks.
A white paper from Apollo Global Management, published in late July 2026, analysed Anthropic enterprise usage data and found “no detectable employment effects” in terms of mass job destruction. Instead, it found that workers in AI-exposed occupations are experiencing a 6.7 percentage point decline in real wage growth since 2023. Employers, the report concluded, are “capturing AI productivity gains through wage compression rather than workforce reduction.”
The impact is not evenly distributed. It is concentrated among the lowest earners.
Who’s Getting Squeezed
The Apollo paper, reported by Axios on July 31, found that service workers face a 24.3% decline in real wage growth, while the bottom wage quartile has seen a 10.7% decline. Top earners show no significant effect.
The most exposed professions include customer service representatives, travel agents, administrative clerks, and telemarketers — roles whose functions are most frequently automated by AI applications. These are not high-paying jobs to begin with.
The total scope is significant. Apollo estimates that 3.7% of the entire US labour force — roughly 5.8 million people — is already experiencing tech-related declines in real income. That figure, the report warns, “almost certainly understates what is coming.”
Forbes reported on August 5 that real wage growth in AI-exposed occupations lagged less-exposed ones by 6.7 percentage points post-2023, translating to an estimated $28 billion annual cost to affected workers.
The Other Side: Skills Still Command a Premium
The wage compression story has a counterweight. PwC’s 2026 Global AI Jobs Barometer, released in June, found that workers with AI skills command a 62% wage premium — up from 57% in 2025. AI-exposed companies have tripled their lead in workforce productivity growth since 2022.
The labour market is splitting in two. On one track, “professionalised” roles where AI acts as a force multiplier for skilled workers are growing twice as fast as other jobs and seeing 42% higher wage growth since 2021. On the other, “democratised” roles where AI lowers the barrier to entry are seeing wages compress.
The Apollo data and the PwC data are not contradictory. They describe the same divergence from different ends.
NZ Angle: Adoption Lags, but the Pattern Will Reach Here
Goldman Sachs research published August 19 placed New Zealand among the developed economies at the lower end of AI adoption, alongside Italy and Japan. Major developed markets have adoption rates of roughly 15% to 20%; NZ sits below that average.
That lag is a double-edged sword. It means the wage compression Apollo describes has not yet hit NZ workers at the same scale. It also means the window to prepare — through reskilling, policy, and collective bargaining — is open now but narrowing.
NZ’s Hays workforce trends report for 2026 notes that 63% of NZ businesses are looking to employ people with AI skills, while 70% report struggling to find them. The demand exists. The supply does not. That gap, if left unaddressed, is precisely the kind of structural condition that produces wage compression for workers without AI skills while inflating pay for those who have them.
The Policy Question
Apollo’s authors were explicit about the implications. The critical question, they wrote, “is not whether AI will reshape the labour market more broadly, but how” — and specifically, how to support “lower-wage workers who have the fewest resources to weather the transition.”
The report’s warning that “more occupations will cross the threshold” as AI adoption deepens is not alarmist. It is descriptive. The 5.8 million workers currently affected are, in Apollo’s framing, “the leading edge of a much larger adjustment.”
For workers, the takeaway is not panic. It is preparation. The data consistently shows that AI skills, communication ability, and judgement are the premium attributes in the emerging labour market. For employers and policymakers, the takeaway is that wage compression is a harder problem to address than unemployment — it is slower, less visible, and easier to ignore.
❓ FAQ
Will AI actually take my job? According to Apollo’s analysis of Anthropic usage data, there is “no detectable employment effect” in terms of mass job destruction. The Stanford Institute for Economic Policy Research found that unemployment for the 20% of workers most exposed to AI rose by 0.77 percentage points since 2022 — less than the 0.85-point increase for the least-exposed workers. The bigger risk, at least for now, is stagnant or declining wages rather than outright job loss.
Which workers are most affected by AI wage compression? Apollo found the effect concentrated among the lowest earners. Service workers face a 24.3% decline in real wage growth, and the bottom wage quartile a 10.7% decline. Customer service reps, travel agents, administrative clerks, and telemarketers are among the most exposed.
How many workers are already affected? Apollo estimates 5.8 million US workers — 3.7% of the labour force — are already experiencing AI-related real income declines. The report says this figure “almost certainly understates” future impact as adoption deepens.
Does AI also create high-paying jobs? Yes. PwC’s 2026 AI Jobs Barometer found a 62% wage premium for workers with AI skills. LinkedIn data shows AI roles paying a median of $177,000, more than double the $80,000 median for non-AI roles. The labour market is bifurcating between AI-skilled workers whose pay is rising and exposed workers whose pay is compressing.
📰 Sources: Apollo Global Management, Axios, Business Insider, Forbes, PwC, Goldman Sachs, Stanford Institute for Economic Policy Research, Hays NZ