video1.0<iframe src="https://www.loom.com/embed/c602d78063304127b972d069ef62c53b" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/c602d78063304127b972d069ef62c53b-b93f13cbbb72bab2.gif4798.033333This Loom discusses how to implement agentic AI to handle support ticket escalations inside Modal, using tools like Cadence and Modal’s MCP capabilities. The team agrees a practical first step is enabling an AI to run tier one escalation work in a human in the loop flow, then iterating based on logs of what the agent needs from humans, with the goal of reducing support load and improving response time and quality. Key constraints include Modal workflow access being limited via current tokens, and the need to manage authorization and access to sub accounts via stored refresh and access tokens on a VPS using a dedicated agency admin email. The discussion also covers the current state of Modal MCP connectors, which exist but are private, and plans to share connector details for a safe demo in a test environment.