Prolific Orchestrates Human-Synthetic Workflows
Jemma White, COO of Prolific, on why humans ensure AI safety
This points to Prolific moving up from selling access to people into deciding how AI teams should combine people and generated data in one workflow. Today Prolific already matches studies to profiled participants, plugs into lab workflows by API, and helps customers find the right cohort fast. Adding orchestration means becoming the control layer that routes each task to synthetic generation, human collection, or human review based on what the model builder is trying to improve.
-
This is how data labeling platforms become more software like. Prolific says frontier labs already use its API for always on access, while self serve and managed services sit on the same system. Orchestration extends that from finding participants to deciding the full path a task should take.
-
The closest comparable is Invisible, which combines software orchestration with a distributed workforce and reached $134M revenue in 2024. Scale is broader still, spanning data labeling, model training, deployment, and synthetic data, which shows how valuable the workflow owner can become versus a pure labor marketplace.
-
For customers, the practical value is cost and speed. Synthetic data is good for producing volume, but Prolific argues human review is still needed for trust, safety, cultural nuance, and specialist judgment. An orchestration layer can choose the cheap fast path first, then insert humans only where model risk or ambiguity is highest.
The market is heading toward mixed pipelines where synthetic data creates breadth and human systems provide the last word on correctness and safety. If Prolific executes well, it can own the routing logic for that pipeline, which would make it harder to replace and push it closer to an infrastructure position in AI evaluation rather than a participant marketplace.