Ooak Data builds RL environments from real company workflows. It sources enterprise documents, communications, and tool data, then anonymizes the material into digital twins intended to retain organizational structure and relationships. The resulting environments contain expert-level, multi-step tasks involving multiple tools, and are calibrated to challenge frontier AI models. Ooak describes the environments as useful for training and evaluating agents on realistic workflows, including for frontier AI labs, enterprise AI teams, and startups. Its public materials also describe multimodal context such as documents, conversations, project-management tools, and organizational charts.