Rachel Woods from AMP explains a framework for building AI playbooks - structured, step-by-step processes that allow AI to perform work reliably without constant human supervision. Each playbook is defined by four components: a trigger that initiates the work, inputs the AI needs before starting, specific steps to follow, and a defined output or deliverable. She demonstrates this with a project planning playbook that, in her company's use, has "tripled the velocity of my team and how we plan and execute projects."
Woods introduces "playbook stacking," where one playbook triggers or collaborates with others to form larger systems. She illustrates this with a sales enablement example: a playbook that answers team questions from a knowledge base in Slack, escalates unknowns to a manager with a proposed answer, and then updates the guide once the manager approves. She also emphasizes "own the playbook, rent the tech," recommending teams store playbooks in GitHub or similar tools rather than locking them inside a single AI platform.
Woods frames a progression from casual AI user to "AI operator" - someone who architects and owns the systems. She advises starting with small, low-stakes playbooks (meeting recaps, weekly reports) before scaling, and notes that the goal is to move from babysitting AI to trusting it to run in the background.