Interventions

Overview

Add human review and approval steps to your workflows

Interventions pause a workflow for human review. Someone examines the data, makes any corrections, and approves or rejects before the workflow continues.

When to Use Interventions

  • Approval required High-value transactions, sensitive data
  • Verify extracted data Confirm AI got it right
  • Exception handling Unusual cases that need human judgment
  • Manual input Information only a person can provide

How It Works

  1. Workflow pauses at the intervention step
  2. Task created in the Interventions queue
  3. Reviewer notified (if assigned)
  4. Review happens approve or reject
  5. Workflow continues with the (possibly corrected) data

Adding to Your Workflow

  1. Drag Intervention from the action library onto your canvas
  2. Configure what data to show for review
  3. Set assignment rules (who handles it)
  4. Connect the next steps

Configuration

SettingPurpose
MessageTell reviewers what to check
VariablesWhich data fields to show
AssignmentWho handles this (user, role, or unassigned)

Handling Interventions

Find Your Interventions

Go to Interventions in the sidebar. Your assigned items appear at the top.

Review the Data

Read the message to understand what needs checking. Review all fields, not just obvious ones.

Make Corrections

Click any field to edit. Fix typos, correct numbers, add missing info.

Take Action

Approve Data is correct (with your changes). Workflow continues.

Reject There's a problem you can't fix. Add a comment explaining why.

Branching After Interventions

Interventions have two outputs: Approved and Rejected. You can route to different steps based on the outcome.

Best Practices

On this page