Emerging New work created by AI deployment. Evidence is younger and thinner.
Data annotators and AI trainers
The most accessible on-ramp into AI work, and the one that rewards domain expertise over coding.
- Horizon
- 2023–2025
- Protection from automation
- Weakly protected · our reading
Every capability claim in this field rests on labour of this kind: people who label, grade and correct model output against a standard. LinkedIn’s platform data puts the title fourth among the fastest-growing roles in the US.LinkedIn 2026
The honest framing: the junior form of this work is precarious hourly gig labour, easily outsourced and likely to be partly automated. The senior form — designing the feedback loops and quality processes for a domain-specific model — is expert work that pays like it. Same title, different jobs, which is the recurring pattern on this site.
What holds it up
- Labeled data and preference feedback are inputs the models cannot generate for themselves
- Domain knowledge (medicine, law, finance) is the scarce input, not programming
What threatens it
- Individual-contributor annotation is hourly, insecure and often outsourced
- Automated labelling compresses the bottom of the market over time