Data Request Agent · Use case
How can we use an AI agent to prepare the answer to an employee's request for their own data on our own infrastructure?
A data request agent takes one employee's request for their own data, lists the systems and files your procedure says to search and what it found in each, and gathers the documents into one set for review. The privacy officer checks the requester's identity, reviews every document, redacts what concerns other people and releases the answer. The agent releases nothing, makes no redaction, decides no exemption and sends nothing to the employee.
One request.
A dozen systems to search.
- ✗ 01An employee's request for their own data has a deadline set by law, and the data sits in the people system, payroll, learning records, case files and mailboxes.
- ✗ 02The set that comes back names other people, and sending it unread is a second problem on top of the first.
Where it stops for a person
Always. The privacy officer reviews, redacts and releases.Always. It gathers and marks, it does not release: what the requester is entitled to, what is withheld and how another person's data is redacted are the privacy officer's decisions, with counsel where needed. A marked passage is a prompt to look, and an unmarked one has not been cleared. A request to erase or correct data is not this agent's job and goes to the privacy officer. What it gathers for one request is not kept as a profile of the person.
It lists and gathers.
The privacy officer reviews, redacts and releases.
Step · 01
The privacy officer confirms who the requester is and opens the request.
Step · 02
The agent reads your procedure for the list of systems to search and what each holds.
Step · 03
It reads that person's records through your own APIs, read only, and lists what it found in each system and what it could not reach.
Step · 04
It gathers the documents into one set, with the source of each, and marks the passages that name or describe another person.
Step · 05
The privacy officer reviews every document, redacts, decides what is withheld and why, and releases the answer.
Reads from · connectors
Your own APIs
The request, as a ticket or an uploaded file; the requester's own records in the systems your procedure lists, read through your own APIs with the operations that change something left out; and your data request procedure, indexed as a knowledge source. It runs with the privacy officer's permissions and reads one person's data for one request.
Acts through · MCP servers
Nothing. It changes and deletes no record, and it sends nothing to the employee or to anyone else.
What you set up
- Index your data request procedure, with the list of systems to search, as a knowledge source.
- Import the descriptions of the APIs that hold employee records and keep only the operations that read.
- Build the agent in Katonic Studio and test it on invented employees, never on a live request.
- Publish it to the privacy team only, and set how long a gathered set is kept after release.
Where it runs
The agent, its knowledge store and the model run on infrastructure you own or control. You choose the model and serve it yourself, so questions, documents and answers are not sent to a public AI service. An employee's records, and the set gathered for a request, stay inside the company until the privacy officer releases the answer.
How you get it today
This page is labelled Use case: the agent is built in Katonic Studio. You describe the job, connect what is named above, add test questions and release the agent once it passes them. When the Catalog holds a tested template for this job, this page will carry the Template label and show how to take it.
Does it send the data to the employee?
No. It has no tool that sends anything. The privacy officer reviews, redacts and releases the answer.
Does it redact other people's data?
It marks the passages that name or describe another person. The privacy officer decides every redaction and reads every document.
Does it handle requests to delete data?
No. It prepares the answer to a request for a copy. A request to erase or correct goes to the privacy officer.