Every enterprise has the system that refuses to die. The one with no API, no modernization budget, and a process owner who knows the click-path by heart. RPA has been duct-taping these workflows together for years, with brittle scripts that break the moment a button moves or a label changes. Computer Use is a different approach. The agent reasons about what it sees on screen and adapts when the screen does not look the way it expected.
This session opens up the playbook. Build a Computer Use agent against a real legacy app, configure the agent for reliable navigation, work through the prompt patterns that separate a flaky demo from a production-ready automation, and handle the failure modes that come with letting an agent drive a real screen. Attendees leave with a working pattern and a list of the workflows in their own org worth automating this way.