JSON multi-field mapping is useful when one reading of the submission should produce several related values.
Example: Support Triage #
Prepare three destination fields:
- AI Summary
- Suggested Department
- Suggested Priority
Create these mapping rows:
| Response name | Destination |
|---|---|
summary |
AI Summary |
department |
Suggested Department |
priority |
Suggested Priority |
Use a System Prompt such as:
Analyze only the information supplied. Do not invent missing facts. Keep summary under 80 words.
Department must be billing, technical, sales, or other. Priority must be low, normal, or urgent.Use a User Prompt such as:
Analyze this support request and produce the configured summary, department, and priority values.
Subject: {Subject:3}
Request: {Request:4}Choose Response Names Carefully #
Names should describe their content and remain stable. They are case-sensitive schema properties, not field labels. Changing a response name changes the structure requested from the provider. Update the prompt when the new name changes what its content means.
Validate Destination Fields #
Use a free-text destination for prose. If a generated value will later control routing, do not assume the AI always produces an allowed choice merely because the prompt asks. Validate or review the result before consequential automation.
Every configured property is required. If one is missing or the response is invalid, the generation is reported as failed rather than silently presenting a partial structured result.
