Summary
Craig Hewitt, who runs a seven-figure software company in about ten hours per week, argues that most small and mid-sized businesses fail with AI because they adopt tools before they understand where work actually bottlenecks. His three-step “leverage sequence” — find the weight, write the handoff, install just one — is designed to fix that ordering problem.
Step one is building a “wait list”: five places where work literally stops and sits in a queue until the business owner personally touches it. This becomes the AI roadmap, replacing the usual model-comparison approach. Step two is writing a handoff document with four components — business context, desired output definition, decision rights (what the AI decides alone versus escalates), and a feedback loop. Hewitt argues that making AI useful is mostly a documentation problem, not a prompting problem, because current models are capable enough that specificity is the binding constraint. Step three is installing exactly one system and proving it works before expanding.
Hewitt illustrates the framework with his own “Decision Partner” document — a two-and-a-half-page business context file he has run 95 real decisions through over five months, including a major pricing change that used to take days but now takes about 30 minutes. The video also references Ahrefs Agent A as an example of a well-scoped, context-rich AI agent. Practical and framework-driven, aimed at founders and operators rather than developers.
📺 Source: Craig Hewitt · Published June 17, 2026
🏷️ Format: Opinion Editorial
