Kana

KANA is a Generative UI application that reimagines how utility consumers interact with their data — moving entirely beyond static dashboards or text-based chat into a fully dynamic, agent-driven interface. At its core, KANA uses an AI agent that interprets each user’s consumption context and responds with live, functional UI components: interactive consumption charts render on demand, savings calculators appear pre-loaded with the user’s actual data, incident report forms surface contextually when anomalies are detected, and exportable report flows trigger automatically based on what the agent determines the user needs.

The stack is built around CopilotKit as the generative UI framework, enabling the agent to render React components dynamically in response to tool calls. The AI layer runs on Gemini, handling natural language understanding, consumption pattern analysis, and savings recommendations grounded in the user’s real data. Data access and tool orchestration are handled via MCP, allowing the agent to call structured tools. The backend is Python-based, processing utility data through an AI-augmented BI pipeline before exposing it as agent-accessible tools.

The originality of KANA lies in applying the Generative UI pattern to a domain — utility consumption in underserved markets — where no intelligent consumer interface currently exists. The agent doesn’t just answer questions; it decides what the user should see, builds it live, and makes it interactive. Every dashboard, every recommendation card, every report flow is a direct response to agent output, making the interface itself the primary expression of intelligence.

no prior work