We are always hesitant to rollout a new tool for users if we are unable to impersonate the agent view for both training and debugging purposes

This would be great because currently it's difficult to create agent facing training materials or validate if an agent is seeing something incorrectly.