X thread

1/10 Delegating more tasks can make the founder bottleneck worse. AI increases the flow of research, code, analysis and drafts. If every uncertain result still returns to the founder, the company has scaled output—not authority. 2/10 A task is not a decision. Task delegation answers: who does the work? Decision delegation answers: who may accept the evidence, make the commitment, handle the exception and reverse the result? 3/10 The bottleneck usually sits at the commitment point: Analysis becomes a budget. A prototype becomes a supported product. Code reaches customers. A draft becomes a public promise. That is where delegated work often returns to the founder. 4/10 Map one recurring decision from request to reversal: Who requests it? Who executes it? Who defines acceptable evidence? Who may commit? What escalates? Who may stop or reverse it? What happens if the approver is unavailable? 5/10 Then write a Decision Boundary Card with six fields: Promise Authority Evidence Exposure Escalation Reversal The unit of delegation is not an isolated task. It is a repeatable decision class with a boundary. 6/10 Promise: what outcome may be committed, and to whom? Authority: name the human owner, the execution delegate and the limit of each role. “The team owns it” is too vague. “The agent decides” does not assign accountability. 7/10 Evidence: what must be true before the owner proceeds? Exposure: what bounds apply to spend, access, customer reach, duration and operational consequence? Authority can usually be wider when exposure is limited and reversal is practical. 8/10 Escalation: which signals move the decision outside the local boundary, and what happens if the escalation owner is unavailable? Reversal: who may pause or roll back, under what trigger, and what record must be preserved? 9/10 The strongest objection is valid: founder centralization can be correct. Keep founder judgment where downside is material, criteria are immature or reversal is difficult. Reduce the active portfolio if the company cannot responsibly govern every initiative. 10/10 The goal is not maximum autonomy. It is the smallest amount of centralized judgment consistent with the promise and exposure. Test one recurring decision: if the founder is unavailable, does it proceed safely, wait by design, or enter an undefined queue?