Rachel Woods Tripled Team Productivity by Treating Claude AI as a Department

An eight-person team — only half of whom work full-time — has tripled its productivity by running AI through structured frameworks called playbooks, according to AI expert Rachel Woods, whose system was detailed in Marketing Against the Grain. Her firm, AMP, uses Claude and other AI tools to automate project planning and sales enablement, with one playbook answering sales questions in one minute where a human would need 10 to 15.

Woods entered AI operations just one month before ChatGPT's release, giving her an early view of the field's rapid shift. Since then, her systems have appeared in Forbes and Business Insider. She calls her method "playbooking," and it rests on a simple premise: AI performs best when it follows a predefined process rather than improvising.

What Is an AI Playbook?

Each playbook contains four components. A trigger starts the workflow — a scheduled task, an incoming email, or an external event such as a change in AI visibility on platforms like ChatGPT. Inputs supply the data and instructions that keep the user in control. Steps outline the process the AI follows, and outputs define the expected results.

The project planning playbook shows the pattern in action. The input is a project pitch. The AI then applies what Woods calls the "131" process:

• Identify the one main problem
• Brainstorm three approaches
• Recommend one

From there, the AI clarifies timelines, creates tasks, and generates a detailed project plan. It also flags tasks it can start immediately, using its own understanding of the project context.

How Fast Do the Results Arrive?

The numbers come from the sales side. AMP's "answering sales questions" playbook pulls from a knowledge base or sales guide and responds to Slack queries within one minute, against 10–15 minutes for a human. Questions the AI cannot resolve get escalated to a manager with a proposed response attached for approval. Once an answer is finalized, a second playbook updates the knowledge base, keeping the information current without manual upkeep.

That playbook also coaches team members on brand communication and researches proposals for managerial sign-off. The team calls the interconnected setup their "go-to-market system" for sales.

Own the Playbooks, Rent the Technology

A core principle of Woods' system is storing playbooks in GitHub — or Google Drive as an alternative — rather than embedding them inside tools like Claude or ChatUp-era chatbots. Owning the playbooks while "renting" the technology gives the team flexibility to swap tools as the field evolves. The team also integrates AI with ClickUp for project management, and while AI handles much of the groundwork, Woods says a robust project management tool remains essential for overseeing complex projects.

The approach extends to personal routines. Woods starts her day by typing "todaying" into Claude, which triggers a playbook that organizes her morning tasks. The team uses Claude's backslash shortcut to locate and run playbooks in new chats, and a playbook called "full pitch" feeds detailed inputs into project planning.

Woods calls this layered technique "playbook stacking": multiple playbooks working together to manage complex workflows, letting AI simulate the work of an entire department with consistent results and less manual intervention.

Not everything runs perfectly. The project planning playbook tends to overestimate task durations — though Woods says refining those estimates is not currently a priority for her company.

Who Builds These Systems?

AMP trains people to become "AI operators" — professionals who design and manage AI systems rather than simply use AI tools. The role involves building systems that align with business goals and deliver sustainable value, and Woods argues it carries significant career opportunity as AI-first workflows become standard across industries. AMP offers a free class on playbooking and a guide titled "Build your first Claude skill file in 30 minutes," which walks users through creating skills in Claude and Codex.

Her sharpest claim concerns instruction quality. Woods argues that most poor AI outputs stem from unclear instructions, not flaws in the technology. The "131" process illustrates her point: by forcing the AI to explore three ideas before committing to one, structured guidance produces better results than an open-ended prompt.

The takeaway for any team weighing AI-first workflows is concrete: define triggers, inputs, steps and outputs; store the playbooks where you control them; and build in feedback loops. As AI-first operations move toward the industry standard Woods predicts, the teams that profit will be those that treat AI not as a chat window but as a process to be designed.