In Plain English
An AI agent in Oracle HCM answers questions from your HCM data and policies, and some can take action, such as updating details or submitting a request. Some only read your data; others can change it.
A Real-World Example
A manager asks an assistant who a new joiner reports to and where they're based, then follows the link it gives to update the joiner's work location.
Why It Matters
Whether an agent only reads data or can change it matters more than its label. It decides how much checking you need before you switch it on.
A Common Misconception
Not every agent works on its own. Oracle has two kinds: assistants you talk to, often called advisors, analysts or concierges, and workflow agents that follow defined steps. Oracle is retiring its older supervisor agent templates in favour of workflow agents.
Practical Takeaway
Before you switch one on, check whose data it can see, whether it can change data, and who owns it.
When You’ll Encounter It
On Oracle HCM pages where an assistant is offered, and in AI Agent Studio, where agents are set up.
What to Look Out For
Access that's wider than the person's normal access, and agents that can change records without a review step.
How to Handle It
Test it in a test environment first, with real roles. Delivered agents generally come with your Fusion subscription, but check your own contract.