Voice Support Agent · Use case
How can we use an AI agent to answer inbound calls on our own infrastructure?
A voice support agent answers inbound calls, works out what the caller needs and answers common requests from your own help content. Anything else is transferred to a person with a summary of the call so far. Speech is turned into text and back by models you serve, and the agent runs on your own infrastructure.
Callers wait.
For answers that are written down.
- ✗ 01Callers queue for answers that are already written down, and staff repeat them all day.
- ✗ 02Call audio is personal data and cannot be sent to a public AI service.
Where it stops for a person
Anything outside the common requests goes to a person.A complaint, a distressed caller, a request it does not recognise and anything that needs an account change go to a person. A caller can ask for a person at any point.
It answers common requests.
The rest go to a person.
Step · 01
Your help content is connected and indexed in the Knowledge Engine.
Step · 02
A call arrives through your telephony platform.
Step · 03
The agent works out what the caller needs.
Step · 04
It answers a common request from your content.
Step · 05
Anything else is transferred to a person with a summary.
Reads from · connectors
SharePoint
Confluence
Your own APIs
The agent connects to SharePoint or Confluence for help content, and to your telephony platform through your own APIs.
Acts through · MCP servers
It answers the caller and transfers the call. It makes no change to an account.
What you set up
- Connect the help content and the telephony platform.
- List the requests it may answer and those it must transfer.
- Test with recorded calls, including ones it should transfer.
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.
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.
Can a caller reach a person?
Yes, at any point in the call.
Which speech models does it use?
Ones you choose and serve yourself.
Where is call audio processed?
On your own infrastructure, with a model you serve. Nothing is sent to a public AI service.