SAP's Synthesis Through Simulation Generates Enterprise Data Without Database Schemas

It is a schema-free data synthesis paradigm in which an LLM agent generates data by executing operations against policy-enforcing APIs inside simulated enterprise environments, guaranteeing structural validity by construction.

Statistical synthesizers rely on seed data and fail without schemas, while STS decouples validity enforcement from distribution modeling, allowing each to be addressed independently without database schemas.

The framework, ten environments, and generated datasets are open-sourced on GitHub, making experimentation immediate, though production adoption requires assessing alignment with local data policies and compliance.

Results are based on ten simulated environments, and schema-privileged agents failed 82% of trajectories in the airline environment, indicating challenges in composing tightly coupled tasks.