This paper explores the groundbreaking potential of autonomous Large Language Model (LLM) agents as Digital Labor in white-collar professions. We focus on Digital Labor’s ability to advance through on-the-job learning from junior to senior roles. Using Bloom’s Taxonomy, we use several white-collar roles to demonstrate how the autonomous agent acquires knowledge about the role and the stakeholders involved. We demonstrate that the agent learns across chat sessions and attentively listens to various aspects such as client profile information, sentiment of the conversation, and customer preferences. The findings of this paper are of significant importance to white-collar jobs, as they demonstrate that Digital Labor can effectively analyze, learn, reflect, and retain knowledge, revolutionizing how we work. The full transcripts of the scenarios can be found at https://github.com/salvella/OTJTraining .