Outside Counsel Invoice Review Agent · Use case
How can we use an AI agent to check law firm invoices against our billing guidelines on our own infrastructure?
An outside counsel invoice review agent reads one invoice from a law firm, sets each line against the engagement terms and your billing guidelines, and lists each line that does not fit, with the rule it cites. The matter owner approves, queries or reduces the invoice. The agent approves, rejects and pays nothing, and it sends nothing to the firm. It runs on your own infrastructure, so invoices and the work they describe are not sent to a public AI service.
Every invoice has hundreds of lines.
The guidelines are read from memory.
- ✗ 01A law firm's invoice lists the work line by line, and the lawyer who approves it has the engagement letter and the billing guidelines in mind, not in front of them.
- ✗ 02The narrative on an invoice describes the advice given and the steps taken on a matter.
Where it stops for a person
Always. The matter owner approves, queries or reduces the invoice.Always. The matter owner decides every invoice. A line on the list is a question for the matter owner, not a finding against the firm: the agent does not judge whether the work was needed or well done, and it gives no score for a firm or a lawyer. A line that is not on the list fits the rules you indexed and nothing more.
It lists the lines that do not fit.
The matter owner decides.
Step · 01
The matter owner or the legal operations team uploads the invoice.
Step · 02
The agent reads each line: who did the work, the date, the time, the rate and the description.
Step · 03
It finds the engagement terms for that matter and the billing guidelines.
Step · 04
It lists each line that does not fit: a rate not in the engagement terms, a person not on the agreed team, a kind of cost the guidelines exclude, a description too short to check, with the rule beside it.
Step · 05
The matter owner reads the list and approves, queries or reduces the invoice, and writes to the firm.
Reads from · connectors
The invoice, as an uploaded file; and the engagement terms for the matter and your billing guidelines, indexed as a knowledge source. It reads no payment system.
Acts through · MCP servers
Nothing. It lists. It approves, rejects, reduces and pays nothing, and sends nothing to the law firm.
What you set up
- Index your billing guidelines and the engagement terms for each matter as a knowledge source.
- Decide which rules are checked: rates, the agreed team, excluded costs, how much a description must say.
- Build the agent in Katonic Studio and test it on invoices your matter owners have already reviewed.
- Publish it to matter owners, one firm's invoices first.
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. Invoices, and the advice their narratives describe, 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 reduce or reject an invoice?
No. It lists the lines that do not fit, with the rule. The matter owner approves, queries or reduces the invoice.
Does it write to the law firm?
No. It has no tool that sends anything. The matter owner writes to the firm.
Where are law firm invoices processed?
On your own infrastructure, with a model you serve. Nothing is sent to a public AI service.