Structured output
Model output constrained to a predefined machine-readable format, such as JSON matching a schema, so downstream software can consume it reliably.
In more detail
A model naturally emits free-form text, which is fine for a human reader and unreliable for a program. Structured output constrains the model to a defined shape, most commonly JSON matching a schema, so the next step in a workflow can read specific fields without fragile text parsing.
This is the seam between the probabilistic part of a system and the deterministic part. The model makes the judgment or extraction; the structured output hands the result to connectors and logic that expect exact fields. Without it, automation built on model output tends to break on formatting variance.
Where this shows up at Ceven
Ceven's AI steps produce structured results where the next step needs them, so a model's decision or extraction flows cleanly into a connector call, a branch, or a record update across the customer's tools. That structure is also what makes each step legible in the audit trail rather than an opaque blob of text.