A profile of Alexandr Wang, CEO of Scale AI, which has 100K+ data labeling contractors, as he looks to expand to higher-margin AI tools to sustain growth
Berber Jin / Wall Street Journal : X: @natolambert , @morqon , @william_fitz , @garybasin , and @deliprao X: Nathan Lambert / @natolambert : Confirmation of one of my points of analysis for the future of data annotation companies: even the best still end up with wild amounts of language model outputs. Prevention/detection is near impossible. Morgan / @morqon : if your data annotation business relies on contracting a small army of labellers, there's a good chance they'll shortcut the work by quietly generating the outputs “100% organic human labelled” is a certification that becomes hard to trust William Fitzgerald / @william_fitz : .@SammBlum recently published story about @alexandr_wang's @scale_AI not paying people. The story was missing crucial info: Who hired Scale AI for the project? Reporting in the @WSJ confirms it was Google. Google screwing over not just workers, but also small biz owners. [image] Gary Basin / @garybasin : In the future, the highest paid (and only) meat job will be organic CoT data producer @deliprao : Multi billion dollar valued startups are doing mistakes students violating GenAI policy will not do
Context & Ripple Effects
Scale AI’s operating model has long combined software with a large distributed labeling workforce: earlier coverage described its Remotasks network at roughly 240,000 workers, extending an even earlier contractor-led image-analysis business. The company’s push beyond labeling is therefore a shift in mix, not a departure from its data-supply role.
The expansion arrives as Scale AI’s commercial and institutional footprint broadens, including a $249M Defense Department contract and reporting that it grew sales sharply in the first half of 2024. Google’s reported use of Scale AI underscores why tooling that sits closer to model-development workflows could be strategically more valuable than labor-intensive annotation alone.
First-order effects
- Scale AI is seeking revenue from higher-margin AI tools, potentially reducing its reliance on contractor-heavy labeling work as its primary growth engine.
- Google and other customers could be offered a broader set of services from the same vendor that supplies training-data work, concentrating more of the development workflow inside Scale AI.
Second-order effects
- Rival data-labeling providers face pressure to add software, evaluation, or workflow products rather than compete chiefly on access to annotators and project execution.
- The value of Scale AI’s contractor network becomes more tied to quality control and integration: allegations around model-generated shortcuts make demonstrable provenance and review processes more important to customers.
Third-order effects
- If annotation vendors consistently move into tooling, data work may evolve from a labor-arbitrage market into a more vertically integrated layer of AI development infrastructure.
- The shift also raises the stakes for auditable human-data claims: as models generate more candidate outputs, buyers may increasingly distinguish vendors by validation systems rather than workforce size alone.
The trend: Data-labeling specialists are moving up the AI stack, pairing human-data operations with higher-margin tools, evaluation, and workflow infrastructure.