How do I set up data parsing and analysis for an AI agent?

Last updated: November 5, 2026

Data parsing lets an AI agent extract structured information from call transcripts, such as outcomes, summaries, categories, numbers, and yes-or-no values. These fields make call review and downstream workflows more consistent.

Before you start

  • Decide which fields you need to extract from calls.

  • Write clear labels and descriptions for each field.

  • Prepare example outputs for fields that need a specific format.

Steps

  1. Open your AI agent.

  2. Go to the Editor tab.

  3. Scroll to Data parsing and analysis.

  4. Click Add field.

  5. Select the information type, such as text, selector, boolean, or number.

  6. Enter the field label.

  7. Add a clear description explaining what the model should extract.

  8. Add an optional format example when the output should follow a specific structure.

  9. Repeat for each field you want to extract.

  10. Drag fields to reorder them if needed.

  11. Choose the parsing model.

  12. Click Save.

  13. Review extracted results in call review under View data.

Tips

  • Use short, readable field labels.

  • Make descriptions specific and unambiguous.

  • Use selector fields when you need consistent categories.

  • Use boolean fields for clear yes-or-no outcomes.

  • Avoid overlapping fields that ask for the same information in different ways.

Troubleshooting

  • Extracted fields are empty: Rewrite descriptions with more specific cues.

  • Wrong values appear: Add examples or clarify the expected output.

  • Selector outputs are inconsistent: Narrow the options and improve field descriptions.

  • Extraction varies across calls: Review the parsing model and field design.