Agents

Agent Data Model

Understand how agent journeys use candidate fields, metadata, and workflow variables during execution.

Definition #

Agent Data Model explains how runtime data is stored and used across a journey, including candidate fields, candidate metadata, and workflow-level variables.

Why it matters #

Most journey issues come from data misunderstandings.
Clear data modeling improves reliability, merge output accuracy, and troubleshooting speed.

How it works #

  • Candidate input is captured at widget level.
  • Values are saved into configured targets (field, metadata, variable).
  • Later steps read those values for routing, rendering, and outcomes.
  • Completion logic uses final state to determine result and follow-up actions.

Configuration fields #

  • Save target per widget response
  • Candidate standard fields mapping
  • Custom metadata keys and labels
  • Workflow variable names and scopes
  • Merge token references
  • Validation and default value rules

Example scenario #

A candidate selects preferred territory in an options step.
That value is saved to metadata, used in a confirmation step via merge fields, and later used by a gate decision for path routing.

Common mistakes #

  • Saving to one target but reading from another
  • Inconsistent naming between metadata key and merge token
  • Missing refresh behavior after update
  • Overloading one variable for multiple meanings

Troubleshooting checklist #

  • Identify expected source of truth for each value
  • Confirm write operation succeeded in backend
  • Verify read operation references the same key
  • Check merge rendering for prompt/display fields
  • Validate path/gate conditions against stored value
  • Re-run flow with clean test data

Last updated Mar 28, 2026