ServiceNow releases AutoSynthData, a training-data generation method for enterprise agents
What happened
On 2 October 2026, ServiceNow CoreAI published AutoSynthData on the Hugging Face Blog, a method for generating training data for enterprise AI agents. The process uses the target model's failures and a stronger teacher's successful trajectories to locate capability gaps, then generates and verifies new tasks for post-training, with the curriculum shifting dynamically towards tasks that remain difficult as the model improves. The process is demonstrated in the EnterpriseOps Gym environment, where tasks consist of a system specification, a user prompt and a verifier, and must meet requirements of feasibility, realism and difficulty.
Written by AI from the reports · updated 14 d ago
Timeline
Follow the reports to see the event from each side.
- Hugging Face BlogAutoSynthData: Generating Training Data for Enterprise Agents
At ServiceNow CoreAI, we built AutoSynthData to turn those capability gaps into training data. It uses a target model's failures and a stronger teacher's successes to decide what the model should learn next, then generates and validates new tasks that exercise those capabilities. As the model improves, the curriculum shifts toward what it still finds difficult.