A new specialist asks what actually determines which Salesforce records an Agentforce agent can see and act on at run time, separate from whatever the pet owner or staff member chatting with it can normally access. What is the correct answer?
Select an answer to reveal the explanation.
Short Explanation
The agent is not wearing your badge. It has its own badge, called the Agent User, and that badge is what actually gets scanned every time the agent tries to open a record or run an action.
Full Explanation
Every Agentforce agent runs in the security context of a dedicated Agent User, and the permission set assigned to that user is what governs which objects, fields and records the agent can read or act on, independent of the human on the other side of the conversation. Org-wide defaults on a single object like Case are one ingredient in the sharing model but are not the mechanism that names the agent's identity, so that option describes a fragment, not the answer. Assuming the agent inherits the logged-in human's profile is the classic and dangerous misconception: it would mean a pet owner's portal login controls what the agent can see, when in fact the Agent User's permission set is what is provisioned and audited by the specialist. There is no such thing as a topic-level access setting divorced from a user record; access control in Salesforce, including for agents, is always anchored to a user and its permission set. Getting the Agent User's permission set right, and not over-provisioning it, is core security configuration work for this role and shows up repeatedly across the AI Agents objectives.