AutoPilot is the hands-on mode of the Resolve agent. It runs on OpenAI's computer-use model, hosted in the platform's Azure tenant. It works in a loop: read the screen, decide the next action, perform it, and verify the result before moving on. When it needs firm knowledge, such as the approved configuration or the right library path, it asks the relevant specialist rather than guessing.
Every action is classified before it runs. Read-only diagnostics proceed. Changes that affect the user or the machine wait for a technician. Actions outside policy are refused and logged.
The POD analyzes the ticket. Resolve fixes the machine. Both learn from the result.