Scam Report Triage Agent · Use case
How can a telecom operator use an AI agent to triage scam reports on its own infrastructure?
A scam report triage agent reads the scam calls and messages that customers report and classifies each one by the operator's own categories. It groups the reports that share a sender, a number or a wording, and drafts the blocking request with its evidence. The fraud team decides what is blocked. The agent blocks no number and no sender.
Thousands of reports.
One campaign.
- ✗ 01Customers forward the same scam in a hundred wordings, and each report arrives as a separate item.
- ✗ 02An analyst who reads them one by one sees the campaign late, after more customers have answered it.
Where it stops for a person
Before anything is blocked.Before anything is blocked. Blocking a legitimate sender harms a business, so the fraud team reads each group and decides.
It groups the reports.
The fraud team blocks.
Step · 01
A fraud analyst hands the agent the reports received since the last review.
Step · 02
It classifies each by your own categories, such as impersonation of a bank or a delivery firm.
Step · 03
It groups the reports that share a sender, a number, a link or a wording.
Step · 04
For each group it writes what the reports have in common and how many there are.
Step · 05
It drafts the blocking request in your own format, with the reports attached. The analyst checks it and decides.
Reads from · connectors
Database
Your own APIs
The reports customers send, in a database with read-only access or through your reporting system's own API. Your scam categories and your blocking procedure are read as a knowledge source.
Acts through · MCP servers
Nothing. It drafts a request. It has no tool that blocks a number, a sender or a link.
What you set up
- Connect the reports with read-only access.
- Write down your categories and what a blocking request must contain.
- Test against a past week whose campaigns you know.
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. Customers' reports and numbers stay inside the operator.
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 block scam numbers?
No. It prepares the request and the evidence. The fraud team decides and blocks through your own process.
Does it read customers' messages?
Only the ones a customer has reported and sent to you. It has no tool that reads traffic on the network.
Can it tell a scam from a real message from a bank?
It classifies by your written categories and shows the reports behind each group. Where it is unsure it says so, and the analyst judges.