Self-serve human data accelerates AI
Jemma White, COO of Prolific, on why humans ensure AI safety
Speed is the real product advantage here, because faster human data collection lets AI teams test prompts, evals, and safety checks on the same day instead of waiting for an ops team to scope the project. Prolific is built so a researcher or model team can set pay, choose a vetted cohort, launch immediately, and get responses back through software driven matching, screening, and fraud checks rather than through manual project management.
-
Managed services usually mean a sales call, a statement of work, custom recruiting, and human QA before any task starts. Prolific still offers hands on support for bespoke pools and complex studies, but its default path is self serve, which removes the slowest step in the workflow.
-
This maps to a broader split in AI data work. Companies like Scale and Invisible grew by wrapping software around large labor operations, while newer platforms like Prolific and Office Hours push more of the matching and qualification into software so customers can start work with less operational overhead.
-
The speed difference matters most when tasks are iterative. Frontier labs and app developers often need to run an eval in the morning, inspect failures, change instructions, and rerun the study that afternoon. A weeks long queue breaks that loop and makes human feedback much less useful.
The market is moving toward hybrid models where software handles the common case and human operators step in only for the hardest projects. That favors platforms that can keep self serve fast while adding expert curation on top, because AI customers increasingly want both immediate turnaround and tighter control over who is doing the work.