Legal Request Triage Agent · Use case
How can we use an AI agent to sort the requests sent to our legal team on our own infrastructure?
A legal request triage agent reads each request the business sends to the legal team, says what kind of matter it is, lists what is missing before a lawyer can start, and proposes who should take it, with the intake rule behind each point. A lawyer or the legal operations lead confirms the kind and the owner. The agent opens no matter, answers no colleague and gives no view on the request. It runs on your own infrastructure, so requests are not sent to a public AI service.
One inbox for every request.
Each is read before anyone knows whose it is.
- ✗ 01Requests reach the legal team by email, chat and a form, and a lawyer reads each one to decide what it is and who should have it.
- ✗ 02A request often arrives without the contract, the counterparty or the date a lawyer needs, and the first reply is a question.
Where it stops for a person
Always, before a matter is opened or a colleague is answered.Always. A lawyer or the legal operations lead confirms every kind and every owner. It sorts a request, it does not answer it: it gives no view on the law, on risk or on what the business should do. A request it cannot place, a request that mentions a dispute, a regulator or an investigation, and anything marked urgent are listed for a lawyer first.
It proposes the kind and the owner.
A lawyer confirms.
Step · 01
A colleague sends a request, or the legal operations lead uploads one.
Step · 02
The agent reads the request and its attachments.
Step · 03
It says what kind of matter it is, by the kinds your intake rules define, and gives the rule it used.
Step · 04
It lists what is missing for that kind of matter and proposes an owner from your allocation rules.
Step · 05
A lawyer or the legal operations lead confirms the kind and the owner, opens the matter in your own system and replies to the colleague.
Reads from · connectors
Your own APIs
The request and its attachments, read as a ticket through your own APIs with the operations that change something left out, or as uploaded files; and your intake and allocation rules, indexed as a knowledge source.
Acts through · MCP servers
Nothing. It classifies, lists and proposes. It opens no matter, assigns no lawyer, changes no ticket and sends no reply.
What you set up
- Write down the kinds of matter, what each needs before work starts, and who takes which.
- Index those intake and allocation rules as a knowledge source, and import the description of the API that holds requests, keeping only the operations that read.
- Build the agent in Katonic Studio and test it on a past month's requests.
- Publish it to the legal operations lead first, one business unit at a time.
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. Requests, with the contracts and names attached to them, stay inside the company.
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 the agent answer the colleague who sent the request?
No. It has no tool that sends anything. A lawyer or the legal operations lead replies.
Does it give a first view on the legal question?
No. It says what kind of matter the request is and what is missing. The view is a lawyer's.
Where are requests processed?
On your own infrastructure, with a model you serve. Nothing is sent to a public AI service.