I built my last venture as a deliberate experiment: how much of a real business can software and AI carry, so a single person’s time goes only where judgment lives? Two constellations made it work — a growth engine that turns creators into conversion, and an operating spine that runs the company. The part I care about most is the boundary: deciding what must stay human.
Discovery to paid distribution: the influencer pipeline that filled the funnel, run on the same one-operator discipline.
A normal day, as designed: the constellation trades events on its own — orders trigger flows, tickets arrive with context and AI-drafted replies, everything reports down into the dashboards. The operator’s queue holds only what deserves a human.
Architecture shown as designed; tools named, numbers and names left out on purpose.
Every order event, support reply, email flow, and report has a system path. The test is simple: if it happens twice the same way, it gets a path and stops costing attention.
The operator’s queue holds three things: exceptions a customer would feel, product and inventory calls, and relationships. Everything else is automation pretending to be work.
Headcount-shaped problems become subscription-shaped. Capital flows into the product instead of the overhead — the principle the whole company was built around.
I design the operating system, not just the process — including the hardest call: deciding what must stay human. The same discipline scales from one operator to a team.