Pay Gap Report Agent · Use case
How can we use an AI agent to prepare our pay gap report on our own infrastructure?
A pay gap report agent reads pay data from a reporting database, read only, works out the pay gap figures by category of worker in the format your reward team has set out from the law that applies to you, and shows the method and the query behind each figure. The reward lead reviews, corrects and publishes the report. The agent reports on groups, leaves out any group too small to hide a person, explains no gap by itself, and proposes no change to anyone's pay.
The report is due each year.
The method is rebuilt each time.
- ✗ 01Pay gap reporting comes round each year, and the figures are rebuilt in spreadsheets by whoever did it last, with the method held in their head.
- ✗ 02Pay data is among the most sensitive a company holds, and a figure for a very small group can show one person's pay.
Where it stops for a person
Always. The reward lead reviews and publishes the report.Always. It works out figures for groups, it does not judge them: why a gap exists, whether it is justified and what is done about it are for the reward lead, management and counsel. It names no person and gives no figure for one. A group below the size you set is left out. It proposes no award and no correction to anyone's pay.
It works out the figures and shows the method.
The reward lead reviews and publishes.
Step · 01
The reward lead names the reporting period and the population.
Step · 02
The agent reads pay, hours, grade and category of worker from the reporting database, read only.
Step · 03
It works out each figure the format asks for, by category of worker, and shows the definition, the query and the count behind it.
Step · 04
It leaves out any group below the size you set and says that it has, and lists records it could not place in a category.
Step · 05
The reward lead checks the figures and the method, writes the explanation and publishes the report.
Reads from · connectors
Pay, hours, grade and category of worker, from a reporting database, read only, through a view your team has prepared; and the reporting rules and definitions, as your reward team and counsel have set them out, indexed as a knowledge source. It keeps to the population the person using it may read.
Acts through · MCP servers
Nothing. It writes to no pay or people system, and it publishes and files nothing.
What you set up
- Prepare a read-only view in the reporting database with the fields the report needs and no more.
- Index the reporting rules and definitions as your reward team and counsel have set them out, and set the smallest group that may be shown.
- Build the agent in Katonic Studio and test it against a report you have already published.
- Publish it to the reward team only.
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. Pay data stays inside the company, and the report is published by a person.
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 show one person's pay?
No. It reports by category of worker and leaves out any group below the size you set. It names no one.
Does it say why there is a gap?
No. It shows the figures and the method. The explanation, and anything done about a gap, are the reward lead's and management's.
Can we check how a figure was worked out?
Yes. Each figure shows its definition, the query and the count of people behind it.