Background on Hackathon Project: Virtual Customer Success Manager (VCSM)

Semantic Kernel is a versatile SDK that seamlessly integrates AI models into applications, enhancing their responsiveness and understanding. It offers tools for chat completion, multi-agent coordination, and data retrieval, enabling developers to create advanced AI-driven solutions.

The Virtual Customer Success Manager (VCSM), developed during Hackathon ’24, is a C# .NET project powered by Semantic Kernel. VCSM transforms customer support, onboarding, and engagement by providing personalized interactions, proactive assistance, and tailored recommendations. By leveraging Semantic Kernel, VCSM adapts to user needs in real-time, enhancing the customer experience and streamlining business processes.

VCSM’s Core Use Cases Powered by Semantic Kernel

VCSM uses three key features of Semantic Kernel (SK) as Chat Completion Model, Multi-Agent Framework, and Vector Store – to drive its core use cases shown below:

1. Personalized Onboarding The Chat Completion Model and Plugins enable real-time, interactive onboarding. Using Automatic Function Calling, VCSM guides users through product setup, automatically invoking functions like GuideInitialSetup() or ExplainFeatures() based on user queries.
2. Proactive Support The Multi-Agent Framework coordinates specialized agents to monitor and resolve system issues. A Master Agent manages agents like the Teams Agent (for Teams performance) and the Azure Agent (for AKS health), ensuring proactive issue resolution and system stability.
3. Tailored Recommendations With the Vector Store, VCSM retrieves customer-specific data such as usage patterns and preferences using the Retrieval Augmented Generation (RAG) approach. This allows VCSM to deliver personalized, real-time recommendations, providing relevant insights based on user history.

By leveraging these SK features, VCSM creates a seamless and personalized experience that improves customer engagement and support.

VCSM Code Breakdown

Here’s a high-level overview of the key modules that implement the onboarding, support, and recommendation features of VCSM. For complete setup and implementation details, visit the GitHub repository. See this link for expected output.

1. VCSM_PersonalizedOnboarding.cs This module manages Personalized Onboarding using Semantic Kernel’s Plugins approach. The TeamsOnboardingPlugin offers an interactive experience by dynamically switching between the following functions based on user requests:
   • GuideInitialSetup(): Helps users with the initial setup of Microsoft Teams.
   • ExplainFeatures(): Explains key Teams features such as chat, meetings, and file sharing.
   • ProductivityTips(): Provides tips and best practices for improving productivity in Teams.

The solution hence adapts to user queries, making onboarding tailored and efficient.

1. VCSM_ProactiveSupportWithAgents.cs This module provides Proactive Support through Semantic Kernel’s Multi-Agent Framework. The Master Agent coordinates responses between:
   • Teams Agent: Monitors Teams performance, resolving issues like disconnections or crashes.
   • Azure Agent: Monitors Azure Kubernetes Service (AKS), detecting issues such as CPU/memory overuse or node exhaustion.

After deliberation between the agents, a final response is crafted and delivered the user based on an agreed Termination Strategy, hence providing real-time support to keep systems healthy.

1. VCSM_TailoredRecommendations.cs This module delivers Tailored Recommendations using the Vector Store to provide personalized insights. Customer-specific content (e.g., PDFs like teamsguide.pdf) is uploaded and used to answer user queries:
   • Profile Setup Instructions: For queries like “How do I set up my profile?”, relevant instructions are retrieved, and a personalized response is generated by the Large Language Model (LLM).
   • Feature Overview: For questions like “What features does Teams offer?”, detailed information is pulled from the Vector Store.
   • Productivity Tips: Personalized productivity tips are extracted from stored content.

VCSM combines user queries with data from the Vector Store to generate context-aware, relevant recommendations that enhance the user experience.

Conclusion

The Virtual Customer Success Manager (VCSM), powered by Semantic Kernel and leveraging Azure OpenAI, shows how AI can transform customer interactions. With features like chat completion, multi-agent coordination, and personalized data retrieval, VCSM delivers dynamic onboarding, proactive support, and custom recommendations. It improves user experience while streamlining business operations, offering a scalable and efficient solution for customer success management.

Thanks to the team for sharing their Hackathon project with us and we’ll continue to feature other projects from last week’s Microsoft Hackathon.