Virtual Assistant Module Overview

The Virtual Assistant Module enables intelligent conversation handling at the start of a call flow, responding to speech input dynamically across various media, including phone calls, webchat, and text interactions.

Functionality & Examples

A Virtual Assistant can be used as an open-question menu, allowing callers to provide natural input for processing.
Example interaction:

Assistant: Good afternoon. How can I help you?
Caller: I have a problem with my rent payment this month.
Assistant: OK. That is a problem. I will put you through to our financial department. They will evaluate the options to solve your problem.

Other types of queries it can handle:

• Business hours inquiries: “What are your opening hours?”
• Service issues: “We do not have warm water anymore.”
• Technical issues: “The electricity is broken.”
• General inquiries: “What will the weather be tomorrow?”
• Availability requests: “Is Patrick Dool available?”
• Integration queries: “Do you integrate with Microsoft Teams?”

Request Analysis & Classification

The Virtual Assistant uses free-form input from callers, structured into predefined topics.

In Artificial Intelligence terminology, these inputs are called intents—in UCS, they are referred to as topics.

Classification methods:

1. Keyword Detection

• Topics are assigned specific keywords that trigger routing decisions.
• If a match is detected, that topic is selected.
• Limitations:
  • Only works for specific keywords.
  • Can lack uniqueness in complex cases.
    ✔ Best for limited topic sets.

2. AI Model (Microsoft Azure)

Ideal for complex classification scenarios.

Created in Microsoft Azure Language Studio with extensive training inputs.

Advantages:

• Distinguishes varied inputs more accurately.
• Improves response accuracy over time. ✔ Maintenance Required:
• Input evolves, so the model must be periodically updated.

Setup Requirements:

• Microsoft Azure Conversational Language API key.
• Configured URL endpoint. ✔ Contact Cirrus Support for assistance.

Directory Assistance

When enabled, the Virtual Assistant can handle directory-based queries, such as:

• “Tom Peters, please.”
• “Connect me with Tom Peters.”
• “Is Tom Peters available?”

✔ No manual topic setup needed—the assistant detects username mentions.
✔ Initiates automatic call transfer to the office or mobile number of the requested user.

Note: Transfers are only available for phone calls.

Using ChatGPT via Azure OpenAI in UCS

Azure OpenAI ChatGPT can be integrated into UCS to provide automated responses based on knowledge base content.

1. Configuration Requirements

✔ Azure OpenAI Account (Create one at Azure OpenAI Studio).
✔ Model Deployment (Contact Cirrus Support for assistance).
✔ Microsoft Azure Conversational Language API key & endpoint (Required for activation).

2. Instruction Prompt Setup

To customise responses, an instruction prompt must be defined.
✔ Example Prompt:

"You are an AI assistant that helps people find information. 
You answer in English with a maximum of 30 words. 
If you don't know the answer, say: Sorry, I don't know. [Help]."

✔ This guides ChatGPT to provide structured responses while maintaining relevance.

3. Response Mechanism

✔ ChatGPT generates answers using knowledge base content and user queries.
✔ Ensures responses are accurate, contextual, and helpful.
✔ Can support FAQ-based interactions within UCS.

4. Support & Implementation

✔ Contact Cirrus Support for help with deployment & configuration.