For most of the past decade, data work — the human labour that trains AI models — meant low pay. An Amazon Mechanical Turk contractor transcribing audio or drawing boxes around cars might earn cents per task. That era is ending from the top: Bloomberg’s 16 October feature documents doctors, lawyers and engineers joining an expert data-work market where a radiologist can earn in an hour what a generalist makes in a week, law firms hire civil-rights attorneys at US$400 per hour to train next-generation AI systems, and LinkedIn ranks data annotator as the fourth-fastest-growing job role in the US.
🔍 THE BOTTOM LINE
The same AI boom that is pressuring entry-level knowledge work is minting a specialist labour market on the training side of the industry. Rates reported by Bloomberg run from US$120 an hour for software engineers to US$170 for condensed-matter physicists and US$400 for civil rights attorneys — and the shift is double-edged: experts are being paid like scarce infrastructure while lower-paid generalists report worsening task availability. For New Zealand professionals, the story is a direct price signal on subject-matter expertise: clinical, legal, engineering and scientific judgement is now an input the AI industry actively buys, on platforms recruiting from LinkedIn profiles.
What the market pays, per the reporting
Bloomberg’s feature opens with two mothers opening laptops for data-work shifts. One, Brook Hansen, a former customer-service worker doing generalist tasks, makes a couple of hundred dollars a week. The other, Anu Brixey, a radiologist, makes the same amount in an hour labelling heart scans on weekends. The gap is the market’s structure in miniature: recruiters hunting specialists on LinkedIn now advertise rates of up to US$120/hour for software engineers, US$170/hour for condensed-matter experts and US$400/hour for civil rights attorneys, while generalist microtask work gets harder to find and slower to pay.
The scale behind those rates is visible in company formation. Bloomberg notes that Mercor — a three-year-old platform recruiting white-collar data workers — has been discussing new funding at a US$20 billion valuation, while Mechanical Turk, the platform that defined the low-wage era, shut down in September after 21 years. LinkedIn’s own annual ranking places data annotators among the fastest-growing US job roles this year.
Why model makers are paying professionals
The economics are straightforward once stated. AI companies have shifted from scraping the open web to buying bespoke, expert-verified data to differentiate their models, and that data now includes clinical scan annotations, legal reasoning traces and engineering calculations. The radiologist in Bloomberg’s story labels calcium in heart scans on weekends partly because she watched an AI tool at her own medical centre produce false positives and concluded that expert judgement would remain part of the loop. MIT professor Danielle Li told Bloomberg the old labour logic — rent expertise, drop it if unpaid — is breaking down because a company that trains a model effectively keeps a version of that labour after the contract ends. That is a plausible reading of why the market is bidding up exactly this kind of work; the counterweight is the generalist tier, where workers describe shrinking task quality and shrinking work.
The two-tier warning inside the good news
This is a hiring boom, but not an evenly distributed one. Bloomberg documents generalists complaining their tasks are harder and steadier work rarer, while specialists command multiples. The lesson for readers is less “data work pays” than “subject-matter depth prices”: the premium attaches to people whose everyday profession gives them exactly the verification skill AI labs need — reading a scan, structuring a legal analysis, checking an engineering result — and nobody should plan a career move on the headline rates without noting how narrow the top tier is.
The NZ angle
New Zealand’s labour market already leans on precisely this structure. SEEK’s September report puts AI skills in 4.2 percent of NZ job ads, nearly double Australia’s 2.4 percent, and the fastest-growing skill categories are agentic AI (up 9.1 percent month on month) and AI ethics and governance (up 8.5 percent). NZ’s AI-skills demand runs at nearly twice the Australian rate — the same specialist premium showing up in local hiring data. Kiwi radiologists, allied health professionals, lawyers and engineers are already exposed to this market through the same global platforms recruiting offshore, and the country’s professional-services base — the groups SEEK shows asking for AI skills most, ICT at 18.8 percent of ads, marketing at 10.4 percent, consulting at 9.5 percent — overlaps almost exactly with the occupations the expert-data platforms target. The realistic local version is not a full-time gig shift: it is a well-credentialled professional adding verified AI-training work at international rates on top of a NZ salary, and the judgement that work exercises is the same one employers inside NZ are starting to pay a premium for.
❓ FAQ
What do AI data-work jobs pay specialists in 2026? Per Bloomberg’s October 2026 reporting: up to US$120 an hour for software engineers, around US$170 for condensed-matter experts, and about US$400 for civil rights attorneys. A radiologist in the piece earns roughly the same amount in an hour that a generalist data worker makes in a week.
Is AI data annotation still growing as a job? LinkedIn ranks data annotator as the fourth-fastest-growing job role in the US for 2026, and specialist recruitment platforms are courting doctors, lawyers and engineers as AI labs buy expert-verified training data.
Did Amazon Mechanical Turk shut down? Yes — Bloomberg reports the platform, which for 21 years defined low-wage microtask data work, shut down in September 2026 as newer specialist-focused platforms such as Mercor (reportedly fundraising at a US$20 billion valuation) grew.
Can New Zealand professionals do this work? The platforms recruit internationally via LinkedIn, so NZ-registered professionals with scarce domain expertise can plausibly access the same rates. Anyone signing up should read platform contracts carefully — the Bloomberg piece does not report uniform engagement terms, and the top-rate tier is narrow.
📰 Sources
- Bloomberg — Doctors and Lawyers Are Joining AI’s $400-an-Hour Gig Economy (16 October 2026)
- LinkedIn — Jobs on the Rise 2026 (fastest-growing US roles)
- SEEK NZ — Employment Report September 2026
- MIT — Danielle Li, labour economics research cited by Bloomberg
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
