For business
What it is
The answer comes back as data with a fixed structure (summary, goals, risks, actions, citations), not free text.
Why it matters
Other systems can use it directly: fill the CRM, generate the PDF, feed a dashboard. Nobody retypes anything.
Risk without it
Free text has to be copied into the right fields by hand, and the format drifts from run to run.
For engineers
How it works
The model is asked for JSON that matches a schema, using native structured-output or tool-call modes where available. The result is validated; on failure the validation errors are fed back for a bounded retry.
Typical stack
- JSON Schema / Zod
- structured-output modes
- retry with error feedback
Pitfalls & trade-offs
- Keep schemas small and explicit: optional fields get skipped.
- Validate business rules, not just types.
- Cap retries and surface failures.
In this example
In
→ draft text
Out
{ summary, goals,
risks, actions,
citations[5] }
schema: valid Fictional sample · not a client system