Approval Workflow

Control what your agents do with Orbitr's human-in-the-loop approval system.

Human Oversight by Default

Orbitr is designed with human oversight as a core principle. By default, agents submit their planned work for your review before executing. This gives you full control while still benefiting from AI-powered research and planning.

How Approval Works

1. Agent Creates Task

An agent identifies an opportunity or action to take and creates a task with a detailed plan.

2. Task Enters Approval Queue

The task appears in your Tasks page with an "Awaiting Approval" status. You receive a notification (if enabled) that new tasks need review.

3. You Review the Plan

Open the task to see exactly what the agent plans to do. This includes:

  • The specific changes that will be made
  • The reasoning behind the recommendation
  • Expected impact on your marketing performance

4. You Decide

For each task, you can:

  • Approve — The agent proceeds with execution exactly as planned
  • Modify — Adjust the plan before approving (e.g., change a headline, adjust a budget)
  • Dismiss — Skip this task entirely. The agent learns from dismissals.

5. Agent Executes

Once approved, the agent executes the task and reports back with deliverables and results.

What Requires Approval

By default, these actions require your approval:

Action TypeExamples
Content publishingBlog posts, page updates, meta tag changes
Campaign launchesNew ad campaigns, email campaigns
Budget changesAd spend adjustments, bid changes above threshold
Structural changesNew pages, URL changes, redirect setup

Notification Options

Configure how you want to be notified about pending approvals:

  • In-app notifications — Badge count in the sidebar
  • Email digest — Daily or weekly summary of pending tasks
  • Real-time alerts — Immediate notification for high-priority items

Best Practices

  • Review regularly — Check your approval queue at least once per day to keep agents productive
  • Start supervised — Begin with full approval mode, then gradually enable autonomy for task types you trust
  • Batch approve — Use bulk actions for groups of similar, low-risk tasks
  • Trust the trends — If an agent's recommendations are consistently good, consider enabling autonomous mode for that task type