{"type":"video","version":"1.0","html":"","height":1440,"width":1920,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1440,"thumbnail_width":1920,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/fe2f52c875e14800acbab807b641e32a-9b7bc56b9f58e95c.gif","duration":968.384,"title":"Team Outdated Journals","description":"The team demonstrated RevReq Wizard, an AI agent built to explain revenue recognition for sales orders by querying product and order data. The presentation covered the problem (customers raise tickets for financial analysis), the system architecture (including deployment to AWS Lambda and use of Cloud Opus 4.6 via AWS Bedrock), a UI demo with chatbot and visualizations, and a deep dive into SSP and ratable plan calculations. Next steps are to expand knowledge coverage to journals and posting workflows, add action-driven capabilities, and integrate additional tools and documentation into the knowledge base.\n\n### Team introduction 0:08\n\n- Aditya Kotikalapudi, Mahrukh Izhar, and Abinandan V introduced themselves and the team name as \"outdated journals.\" \n- The presenters confirmed presence and readiness to begin the presentation before moving to the demo.\n\n### Problem statement 1:08\n\n- The presenters explained that many customers raise tickets to request financial analysis and to understand how revenue is being recognized for sales orders. \n- The team stated the objective of introducing an agent so users can chat with a bot to learn about order revenue and to increase transparency into internal calculations.\n\n### Solution and architecture 2:00\n\n- The team described the AI agent named RevReq Wizard as a multi-tool agent that has access to APIs containing the required information. \n- The backend architecture was described as deploying the agent to AWS Lambda and connecting with AWS Bedrock while using Cloud Opus 4.6 as the model. \n- The presenters specified that the knowledge base uses product documentation from Confluence and ChargeBee docs and that the Sales Order API is used to fetch details about a given sales order. \n- The frontend was described as a React UI widget that renders real-time charts using Apache Echarts.\n\n### Planned next steps 3:13\n\n- The team plans to expand knowledge coverage and tool coverage to include journals and explanations of how posting works. \n- The presenters intend to add action-driven capabilities that can update sales orders and contract date values and perform actions on behalf of users. \n- The team also proposed integrating more tools and adding more examples and complex recognition rules and SSP configurations into the knowledge base by incorporating Confluence and SSP documentation.\n\n### Demo and deep dive 4:01\n\n- The presenters showed a single-page UI with a revenue analyzer button that opens a chatbot widget where users can ask questions about orders. \n- The bot was demonstrated fetching order data (for example, SO1) using the Sales Order API and returning comprehensive explanations of rules, allocation logic, and revenue plan schedules. \n- The presenters showed a bar chart that explains revenue across each period for an order and mentioned that the UI can visualize the calculation as a graph and provide a detailed chat view. \n- The demo included a breakdown of products, their SSP configurations, allocated revenue per order item, and an added column labeled \"how it is calculated\" to show internal calculation logic so users do not need to rely on external spreadsheets. \n- The presenters explained ratable plan details, including a daily rate calculation presented per month with differing month day counts and a total sum, and they showed that the model will interpret chart and context history to answer deeper questions. \n- The team identified that the agent can reduce support dependency for job errors by enabling explanation and analysis through the agent."}