An Assistant is the core building block of Ringg AI. It is an AI-powered voice agent that conducts phone conversations on your behalf, following a defined objective, script, and voice persona.
Assistant Types
By Orchestration
| Type | Description | Best For |
|---|
| Single Prompt | One unified prompt drives the entire conversation | Simple, linear workflows (e.g. reminders, quick surveys) |
| Multi Prompt | Multiple prompt nodes form a branching conversation flow | Complex workflows with conditional paths (e.g. loan collections, onboarding) |
By Call Direction
| Direction | Description |
|---|
| Outbound | The assistant dials the contact’s number |
| Inbound | The assistant answers calls made to your number |
| Webcall | The assistant handles calls initiated from a web widget embedded on your site |
Industry Templates
When creating an assistant, you can choose from pre-built industry templates:
Templates pre-populate prompt structures suited to that industry, saving setup time.
Assistant Settings
Every assistant has the following configurable settings:
Assistant Settings
| Setting | Description |
|---|
| Prompt | The first message the assistant speaks, the conversation objective, and response guidelines |
| Voice | Language (primary + secondary) and voice character (Tanya, Naina, Raj, Rahul, Anya, etc.) |
| Custom Variables | Dynamic variables injected into the prompt at call time (e.g. {{callee_name}}, {{mrp}}) |
| Knowledge Base | Attach a Knowledge Base so the assistant can look up information during calls |
Call Settings
| Setting | Description |
|---|
| Call | Maximum call duration, inactivity timeout, timezone, noise filtering, voicemail detection, retry logic, and audio behaviour |
| Chat | Session duration and idle timeout settings for webcall/chat interactions |
Advanced Settings
| Setting | Description |
|---|
| Embed | Widget configuration and auto-generated embed code for webcall integration |
| Keyword Boosting | Add domain-specific words to improve the assistant’s speech recognition accuracy |
| Custom Analysis | Define a post-call analysis prompt and output schema to extract structured data from each call |
The Prompt Structure
A well-structured prompt typically contains:
First Message
The exact opening line the assistant speaks when the call connects. Supports @{{variable}} syntax for dynamic personalisation.
Objective
A paragraph describing the assistant’s role, context, and goals.
Response Guidelines
Rules for tone, persona, length of responses, and language switching.
The Interruptible toggle on the first message controls whether the callee can speak before the assistant finishes its opening line.
Agent ID
Every assistant gets a unique Agent ID (UUID). This is used for API calls and integrations.
Test Call
Use the Test call button (top-right of any assistant page) to trigger a live test call before deploying to a campaign. 
