AI review: Autopilot recommendations

How Autopilot recommendations are made, what they mean, and why a person always decides.

Autopilot looks at open HR work items and suggests what to do next, such as review, escalate or complete. Its recommendations are rule-based: they come from fixed rules about each item’s due date, SLA, category and workflow status. They are not produced by machine learning or a generative AI model. Each recommendation shows the model label rules-v1.

Each recommendation comes with a risk level, a confidence and its reasoning, so you can see why it was suggested. The confidence reflects which rule matched — for example, overdue items get a higher confidence that they need attention.

1. Go to Inbox → AI review.
2. Select Refresh analysis to update recommendations for open items.
3. Open an item to read the Autopilot recommendation, its confidence and its reasoning.
4. Select Accept guidance or Dismiss. Either way, the item is marked Human reviewed.
5. Take the actual action on the item yourself.

Autopilot is guarded. No employee or payroll decision is applied automatically — nothing consequential happens until a person reviews it. Only owners, administrators and HR can run it.