The Governance Spectrum

The Governance Spectrum: Autonomy vs Approval-by-Default

Every Paperclip company sits somewhere on a spectrum between two extremes: a board that approves every action before it happens, and a board that approves nothing and lets agents run completely free. Neither extreme works, and the right position depends on how well you know your agents, how consequential their mistakes are, and how much of your own time you're willing to spend reviewing.

This guide explains the spectrum, Paperclip's default governance model, the four approval types you can use, and the practical steps for tightening or loosening governance as your company matures.

---

The spectrum

Full approval Full autonomy | | | ← approve everything approve nothing → | | | [Operator reviews] ──────────────────────────────── [Agents run free] every action strategy + hires only no gates at all

At the left end, the board approves every task transition, every hire, every expenditure. The company never takes a consequential step without a human in the loop. Nothing goes wrong that you didn't sanction — but the agents are nearly useless because they're waiting on you all the time. You've hired an AI company and put yourself as the bottleneck at every step.

At the right end, agents hire agents, spend freely, set their own strategy, and ship work without anyone reviewing it. They're fast — and when something goes wrong (the CEO interprets the goal too literally, an engineer ships a destructive migration, the CMO pursues a brand direction you never wanted) there's no mechanism to catch it before damage is done.

The useful range is everything in between.

---