The AI-native PM approach makes sense for teams already comfortable with conversational interfaces. The ability to bring in specialized agents for different domains could reduce context switching between tools. How does the agent coordination work when multiple specialists are involved in the same project? Like if the marketing agent creates requirements that affect the coding agent's tasks, does the system handle those dependencies automatically or require manual oversight? Also curious about the third-party agent integration. What's the approval and security model for bringing external agents into project workflows that might contain sensitive business information?