Summary
Creator Sharbel A. makes the case that most people misuse AI by treating it as an on-demand chatbot rather than designing role-based “AI employees” that own entire repeatable workflows. He walks through his personal roster of named agents—Nova (YouTube content strategy pulling from Notion, VidIQ, and competitor signals), Sage (Twitter/X posting), a React employee (time-sensitive news packaging), an email partnership agent, a client pulse monitor, and research assistants—each built with specific standard operating procedures, approval gates, and feedback loops.
The video’s core argument is that the prompt is just the “job interview”; the real leverage comes from defining a role, granting the right tools and memory, setting hard limits on autonomous action, and building an improvement loop over time. Sharbel draws a sharp line between “using AI” (ask, get answer, repeat) and “managing AI” (define ownership, context, and boundaries upfront). He also notes that Nova is open-sourced and available for download.
The second half provides a structured template for designing a first AI employee: identify a recurring workflow, define the exact output the agent owns, specify what information it needs, schedule when it runs, set approval gates, and establish a report format. Practical examples are given for creators, agency owners, and salespeople, making this a useful entry point for anyone moving from ad hoc AI prompting toward systematic agent design.
📺 Source: Sharbel A. · Published July 15, 2026
🏷️ Format: Workflow Case Study
