AI Copilot

Introduction to AI Copilot

AI Copilot is your all-in-one conversational AI assistant, designed to simplify and enhance the way you build, manage, and optimize AI agents. It combines automation, adaptability, and transparency to create seamless, high-quality conversational experiences. CoPilot ensures your AI agents deliver consistent, engaging, and effective responses, whether addressing customer inquiries, refining workflows, or optimizing small talk and knowledge base interactions,

CoPilot offers a user-friendly interface and robust capabilities and enables you to:

• Test and validate structured conversations to ensure they align with your specific goals and use cases.
• Enhance small talk interactions for more natural, relatable, and engaging conversations.
• Optimize knowledge base responses, ensuring accurate and contextually relevant answers for every query.
• Monitor and refine interactions in real time, gaining valuable insights into AI behavior and performance.

How AI Copilot Works

1. Onboarding and initial setup

• Agents configure bots to meet essential goals and use cases, such as workflows, small talk, and knowledge base integrations.
• The onboarding process is intuitive and guided, providing a step-by-step walkthrough to set up and fine-tune your bot.
• After initial setup, CoPilot becomes an interactive assistant, enabling you to test, refine, and manage bots effortlessly.
• A conversational interface offers detailed insights into each conversation flow, helping you monitor performance and optimize responses.

2. Real-time interaction monitoring

• CoPilot tracks user interactions in real time, categorizing each query into distinct types:
  • Small Talk: Handles casual interactions like greetings or icebreakers.
  • Knowledge Base (KB): Responds to queries using indexed documents, FAQs, or external links.
  • Generic Queries: Uses advanced language models to address broader, less structured questions.
• Every response includes detailed reasoning, giving you clarity into the decision-making process.

3. Testing & refinement

• CoPilot provides a unified interface for testing conversation prompts, workflows, and configurations.
• You can instantly test changes without switching contexts, ensuring faster iteration and optimization.
• Each conversation trace includes components like prompts, configurations, and token usage to help you understand and improve system behavior.

---

Features of AI Copilot

1. Real-Time Testing and Modification

• Test and update conversation flows, small talk, and knowledge base configurations in real time.
• Quickly implement and test changes within the same interface, reducing time spent switching tools or workflows.

2. Detailed Insights and Monitoring

• Access logs, traces, and reasoning for every interaction to better understand how the system processes and generates responses.
• Use categorized insights to identify opportunities for improvement or fine-tuning.

3. Dynamic Configuration Adjustments

• Modify AI configurations, prompts, and workflows dynamically to adapt to evolving use cases or business requirements.
• Ensure your bot remains relevant and effective, even as objectives change.

4. Multi-Modal Response Handling

AI Copilot can seamlessly manage different types of user interactions, ensuring tailored responses for every context:

• Small Talk: Keeps users engaged with natural, conversational responses to casual queries.
• Knowledge Base (KB): Delivers precise answers from indexed resources, ensuring accuracy and relevance.
• Conversations: Handles goal-driven interactions, such as booking appointments or processing orders, by following predefined flows.
• Generic Queries: Leverages Large Language Models (LLMs) to address broad or unstructured questions, expanding the bot's versatility.

---

How to use Copilot

Enter your intent

1. In the Cloud Platform, navigate to Automation > Copilot.

   

2. Enter your intent in the input chat field. Only intents you’ve added to your conversation, small talk, or questions related to knowledge base content are supported.

3. You can view detailed insights into the conversation in the Traces section. Click Show Trace to access it. Each trace includes components that explain how the system processes user inputs, determines actions, and generates responses.

Traces provide all the generated settings derived from the conversation's context and you can fine-tune them if needed to ensure the agent accurately fulfills the user's requirements.

What Action Could Be Taken

This section dynamically lists all the possible actions generated by the Copilot based on the specific use case and context of the conversation. You can fine tune it if needed.

Option	Description
Upload document	Suggests uploading a file or document if relevant to the conversation context. Example: If the user asks to analyze a report or verify a resume, the agent prompts for a file upload.
Edit trigger	Helps you to refine or modify the trigger conditions that activate a conversation or action within the bot. This ensures that the bot responds appropriately to user queries by aligning triggers with your intended use cases.
Edit conversation prompt	Allows you to refine how the agent interprets and responds to the user’s query. Example: If the AI agent's response is off-target, you can edit the prompt to improve clarity and alignment with user intent.

Customizations: Revise the phrasing for better comprehension.; Add details or context to guide the agent toward a more accurate response.

Agent Response: Conversation

This section provides a detailed breakdown of how the AI agent processes user inputs and generates responses. This trace enables you to understand the flow of communication and refine the agent’s behavior as needed.

The following are the components of this section:

Option	Description
User message received	Displays the original message entered. Example: "Book a demo."
Agent prompt	Shows the generated prompt that guidest the agent’s response based on the query.
Agent config	Contains the configuration details applied to the conversation. This might include api version, models, AI used, or any predefined parameters.
Agent history	Tracks the sequence of interactions in the current conversation. You can verify if the responses maintain continuity and preserve context.
Agent response	Conversation. Displays the agent’s response generated for the user’s message.
Agent reasoning	Explains the logic used to determine the agent’s response, offering insights into how decisions were made.
Predicted tools	Lists any tools or APIs predicted to be necessary for completing the task.
Token usage	Provides details (JSON) about the number of tokens used during the conversation, helping to monitor performance and efficiency.

Conversation

This section provides insights into the conversation flow, whether it’s a predefined conversation, a knowledge base response, or another type of interaction. It helps you analyze and refine the conversation to align with the user's intent.

Option	Description
Conversation prompt	Displays the initial prompt generated for this specific conversation.
Conversation config	Displays the configuration settings for the current conversation, such as parameters, logic, or flow control.
Conversation history	Tracks the sequence of exchanges within the conversation, including user queries and agent responses.
Token usage	Monitors the number of tokens used for the conversation, helping to manage resource consumption effectively.
Conversation response: Message	Details the agent’s response (JSON) generated for the user’s query in the conversation.
Conversation reasoning	Explains the logic behind the agent’s response, detailing why a specific reply was chosen based on the context.