Invisible as AWS for labor
Invisible at $134M in revenue
Invisible’s real play is to turn outsourced work into programmable infrastructure, which makes it look less like a staffing firm and more like a layer companies plug into when software alone still breaks. The key is not just access to 3,000 plus workers, but software that splits a messy job into steps, routes each step to the right human or bot, and sells the result as a managed workflow for AI training, onboarding, data cleanup, and other back office work.
-
In practice, an AI lab can send thousands of model outputs to Invisible for rating, ranking, and safety review. Invisible charges roughly $30 to $45 per hour for this work, pays raters about $15 to $20 per hour, and uses its internal workflow system to divide the job into repeatable microtasks.
-
That puts Invisible between Scale AI and a classic BPO. Scale is more API and dataset driven. Accenture, TaskUs, and Teleperformance sell larger fixed teams on longer contracts. Invisible’s edge is on demand labor wrapped in workflow software, which lets it jump from RLHF to menu onboarding or customer operations without rebuilding the whole system.
-
Mercor shows the next frontier of this market. As RLHF shifts from low cost generalists to $50 to $100 per hour doctors, lawyers, and PhDs, the winning vendors will be the ones that can source talent, test quality, and manage workflows in regulated domains where every judgment needs to be traceable.
The market is moving toward narrower, higher stakes workflows where humans stay in the loop because the cost of a wrong answer is too high. If Invisible keeps turning labor into modular products, it can expand from AI labeling into healthcare, finance, defense, and enterprise operations as the orchestration layer for work that AI still cannot safely do alone.