How Luma VA Works

Luma Virtual Agent (Luma VA) is an advanced support agent designed to assist end-users by understanding and processing their requests using advanced AI technology. With the power of large language models (LLM) and Generative AI, Luma VA can understand what users are asking for and provide the right information or assistance. It identifies user intent, understands their emotions, and offers relevant solutions, making the support experience smooth and efficient. This page explains how Luma VA works and what you can expect.

Initiating a Conversation

When an end-user initiates a conversation with Luma VA, the system uses Generative AI to comprehend the user's request. This involves identifying the user's intent and responding appropriately. Here’s a step-by-step breakdown of how Luma VA processes user interactions:

Open

1. Intent Identification: Luma VA uses advanced AI to determine the intent behind the user's request by analyzing the language used. It identifies one of the following intents:

   • Small Talk: Casual or non-task-specific conversation.

   • Seeking Knowledge: The user is looking for information.

   • Reporting a Problem: The user is reporting an issue.

   • Requesting a Service: The user needs a service.

   • Looking for Ticket Data: The user is searching for information on their existing tickets.

   By accurately identifying the intent, Luma VA ensures that each request is handled appropriately and efficiently.
   
   

2. Sentiment and Tone Analysis: Luma VA can detect the sentiment and tone of the user's message. By analyzing these elements, Luma VA adjusts its responses based on the user's emotions, ensuring a more empathetic and appropriate interaction. This ability enhances user satisfaction by making the interaction feel more personalized, helps in de-escalating potentially tense situations, and provides support that is sensitive to the user's current state.
   
   

3. Categorizing Requests: Based on the identified intent, Luma VA categorizes the request into one of the following:

   • Problem: These requests involve IT-related support issues, such as hardware failures, software errors, network disruptions, security incidents, service unavailability, data loss, user errors, etc. Non-IT-related issues may include facility-related problems, access control and security issues, general inquiries, telecommunications issues, equipment malfunctions, office supplies requests, travel arrangements, employee onboarding and offboarding, and more.

   • Data Request: The request for raw data, facts, or specific details, such as ticket information or details from any system.

   • Knowledge: This is a request where the user seeks precise instructions or solutions to specific topics. These may include requests for Knowledge Articles, step-by-step instructions, troubleshooting guides, frequently asked questions (FAQs), and other detailed information.

   • Service Request: This option is used when the user wants to request a service or place an order. It covers a wide range of requests, including IT-related support for changes in infrastructure, systems, applications, or other IT components, as well as non-IT services.

   • Ambiguous: This category is used when the user's request cannot be definitively deduced or interpreted from the text, necessitating further clarification to eliminate uncertainties. It applies to vague, unclear, or difficult-to-understand requests that require additional information to establish clarity.

   This categorization ensures that each request is directed to the appropriate process, allowing for efficient and accurate handling of user needs.

Handling Different Request Types

Open

Problem Report

When a user reports a problem:

• Luma VA searches the internal Knowledge Base for pre-configured services and relevant articles.

• The information is presented based on how confident the system is about the found information.

• If information is not found in the internal Knowledge Base, the system is not confident in the search results, or the user does not like the results, fallback is triggered.

• Luma now gathers detailed information about the issue and uses it to log a well-formed ticket, ensuring that analysts have all the necessary details to resolve the problem efficiently.

Service Request

When a user needs a service:

• Luma VA searches the internal Knowledge Base for pre-configured services and relevant articles.

• The information is presented based on how confident the system is about the found information.

• If information is not found in the internal Knowledge Base, the system is not confident in the search results, or the user does not like the results, fallback is triggered.

• Luma VA collects the required information from the user and submits the service request to the appropriate team for action.

Knowledge Request

When a user seeks information:

• Luma VA searches the internal Knowledge Base for relevant articles and presents the information to the user.

• Luma searches the internet for relevant information if the information is not found.

Ticket Data Search

When a user searches for ticket information:

• Luma VA uses Natural Language Processing (NLP) to understand and process the query.

• It retrieves and presents relevant ticket data to the user.

• User can now take further actions on their ticket based on ITSM practices configured in their system.

Ambiguous Requests

When the user’s request is ambiugous and system cannot understand the intent,

• Luma prompts the user to rephrase or provide more details on their query.

• In case, an unknown acronym is used in the phrase, Luma prompts the user to confirm its understanding.

• System categorizes the request again and presents the information according to the new intent and category.

Managing Feedback

Luma VA captures and collects feedback from every user interaction. This feedback is essential for improving the system's accuracy and performance. Here’s how feedback is managed:

• Knowledge Gaps: When Luma VA cannot find relevant information in the internal Knowledge Base, it implicitly logs this as a Knowledge Gap. Knowledge Curators and Administrators can use this information to create new Knowledge Articles and configure new Services.

• Positive Feedback: Positive feedback helps identify the most helpful content. It assists curators in understanding the type of information users seek and ensures that the most helpful articles are highlighted.

• Negative Feedback: Negative feedback is marked against the specific Knowledge Article served. Curators use this feedback to update and improve the content of the Knowledge Articles, ensuring that future users receive more accurate and relevant information.

• Key Performance Indicators (KPIs): The feedback collected is used to update Key Performance Indicators (KPIs), which reflect the overall service quality. Continuous feedback allows Luma VA to adapt and enhance its support capabilities.

Luma is designed to streamline support processes by efficiently understanding and handling user requests. Leveraging Generative AI, sentiment analysis, and NLP, Luma VA ensures that users receive timely and relevant assistance. The continuous feedback collection and intelligent fallback mechanisms enhance the user experience, making Luma VA a robust and adaptive virtual support agent.