Practical essays on company context, AI agents, workflow design, and the work of making an owner-operated business less dependent on one person’s memory.
Automations, skills, and agents are three different ways to use AI to get work done inside a business. The useful distinction is how much context, judgment, and ongoing responsibility each one is being asked to carry.
AI agents become useful only after the business becomes legible to them. A company context library gives people and AI a shared, access-controlled place to understand how the business works, makes decisions, and judges good work.
AI autonomy should expand from evidence, not optimism. Before a skill or agent gets more room, the business needs a record of reviewed work, clear standards, useful evals, and a way to pull the freedom back when the work drifts.
SOP libraries go stale when they are maintained outside the work they are supposed to describe. Useful company knowledge needs ownership, review, and a way for new lessons from daily work to become official without turning into a documentation chore.
AI agents can prepare decisions, gather context, draft recommendations, and route work to the right person. The line is final authority: decisions that change promises, policy, money, company knowledge, or accountability should stay with people.
The first AI workflow should be a narrow, reviewable piece of repeated work: one where AI can gather context, prepare a useful output, and hold judgment or customer-facing action for a person to review.