Limit agents to the requesting employee’s own records where appropriate.
A benefits question resolves against the asker’s own eligibility data — the agent never needs, or gets, anyone else’s file.
Let agents support employees without exposing sensitive workforce data or making unauthorized decisions.
HR agents frequently need access to highly sensitive records: compensation, health information, performance reviews, immigration data, investigations, and organizational changes.
people-agent, acting for the employee asking about their own benefits. The held fields are not redacted after the fact — they are never inside the request.
A benefits question resolves against the asker’s own eligibility data — the agent never needs, or gets, anyone else’s file.
The agent’s reach mirrors the human role it serves, so a manager’s assistant sees a manager’s slice and nothing more.
Each sensitive category is its own scope; access to one implies nothing about the others.
People-impacting write actions are not given to the agent, so they never reach the tool list it is handed, no matter how routine the request appears.
Coordination runs on the data the process needs, without exposing unrelated candidates or internal records.
Access follows the relationship and ends automatically, instead of sitting on someone’s cleanup list.
Every status, pay, or access change carries who asked, which agent acted, and the verdict on the call.
Explore the context attached to every request.
people-agent has a defined purpose and a bounded set of tools.
The request acts for hr partner, using that person’s permissions.
The policy applies to the requested action within Employee records, before the tool executes.
Scoping to the requester’s own record makes “what’s my coverage” safe to automate at scale.
Scheduling data flows freely while other candidates’ records stay outside the agent’s scope.
Provisioning follows the role’s approved resource list; privileged grants are not in what the agent holds, so those stay with IT or HR.
Drafting uses the manager’s own team data; another team’s reviews are simply unreachable.
Yes. It drops into the stack you already run, inheriting roles and relationships from your identity systems. The HRIS stays the system of record; Fabriq decides, per request, which records an agent can touch and which actions it may take at all.
Recruiting agents are limited to the candidates and interviewers in their process. They can coordinate schedules and communications without the ability to reveal unrelated candidate records, and their access ends when the recruiting relationship does.
Yes — grant the reads and leave the writes out, per person and per action. A change the agent was not granted never appears in its tool list, so the step stays with an HR partner.
Yes — per user. An employee’s assistant resolves against their own records; a manager’s assistant widens only to that manager’s team; sensitive categories like compensation and investigations stay separately gated regardless of who is asking.
Revoke the person’s session and it dies immediately; their agents lose access with them, and the history of what those agents did stays reviewable.
Every action is attributed — the employee who asked, the agent that acted, and the verdict — so questions about access to people data end in a query, not an interview.