Per-Unit Pricing Fuels Scale Growth
Scale AI
Usage pricing turns better tooling directly into more billable output, which is why Scale can grow faster than a classic outsourcing vendor even though both depend on human labor. When Scale improves pre-labeling, routing, or QA, each contractor can finish more images, video clips, or RLHF tasks per hour, and customers are still charged per completed unit or committed volume, not per seat or hour. That lets software gains show up as revenue growth and margin expansion at the same time.
-
A traditional BPO usually sells labor time. If workers get faster, the client often buys fewer hours for the same job. Scale sells finished work units, like images, annotations, and model evaluation tasks, so faster throughput increases the number of units processed and billed instead of shrinking the invoice.
-
The model also makes buying easier. Self serve per unit pricing gives an ML team a simple cost formula before procurement starts, while enterprise contracts layer in annual volume commitments and discounts. That is much closer to cloud infrastructure buying than to negotiating a staffing contract with an outsourcer.
-
This is part of why software assisted labor businesses can look structurally stronger than older labeling vendors. Scale has been estimated at 50%+ gross margin, above generalized labeling and outsourcing peers like Appen at 24% gross margin and TELUS at 21%, because its software increases output per worker while pricing stays tied to customer usage.
Going forward, the same pricing advantage should keep rewarding Scale as annotation shifts from raw labeling to higher value model evaluation and expert feedback. The winners in AI data work will be the firms whose software keeps raising output per worker while preserving per task pricing, because that is how a labor business starts to compound more like infrastructure.