Scaling a Startup With AI Playbooks Instead of New Hires

Woods, a former Facebook AI researcher, advocates for a "playbooking" methodology to keep operations running without constant human oversight. The process begins by breaking down repetitive business tasks into granular steps—much like training a new hire. By documenting the inputs, decision-making criteria, and desired outputs, the team creates a blueprint that allows AI agents to execute work independently. If the output misses the mark, the team iterates on the playbook logic rather than the individual result, ensuring the system becomes increasingly reliable.

At AMP, these playbooks handle critical tasks ranging from reviewing sales calls to answering internal team queries, effectively acting as an automated operations coordinator. While the company relies on subscriptions to tools like Claude and ChatGPT, the bulk of their operational cost comes from background execution. In September alone, the firm spent $8,510 on Claude Code to facilitate these autonomous processes. Woods emphasizes that the key to avoiding "rogue" AI is maintaining human ownership of the underlying logic. By focusing on the strategy and "renting" the underlying technology—whether it be Microsoft Copilot, Gemini, or Grok—the team remains platform-agnostic, ensuring their business processes survive long after any single AI model evolves.