Atlassian's Jira Service Management (JSM) has developed a Virtual Agent to enhance customer support through automation. This AI-driven solution enables faster, more accurate responses to customer inquiries across various channels like email, Slack, and Microsoft Teams. Customers can initiate requests through different platforms, and if the Virtual Agent can't resolve the issue, it seamlessly escalates to a human agent while creating a ticket. The architecture of the chat system has evolved significantly. Initially, JSM faced challenges with inconsistent bot responses across channels and had to manage six different backends. The new architecture categorizes channels into streaming and non-streaming types, allowing for real-time responses in the former. A unified orchestrator now ensures consistent replies, and conversation data is securely stored in DynamoDB. This overhaul has improved the overall efficiency of the system. Key features of the Virtual Agent include a straightforward routing strategy that checks for intent matches before escalating to human agents. The system personalizes user queries, utilizes a Retrieval-Augmented Generation (RAG) approach, and implements safeguards to prevent misleading AI responses. The impact has been significant: there's been a 50% increase in resolution rates through automation and a 40% boost in customer satisfaction scores. Support is now available in over 20 languages, making it accessible to a wider audience. Enhancements like query variation and a hallucination detection mechanism further improve the quality and reliability of responses.