Tiered Authorship

Decouples who may raise demand from who may cause code to run, so the safe population can be very large and the risky one very small.

Authorship is not a single permission. It is tiered, and granted per person.

  • Tier 0 contributes ideas. External analysts default here, siloed by client organisation.
  • Tier 1 collaborates on specifications.
  • Tier 2 may cause code to execute.

Why the tiers exist

Treating “can raise a suggestion” and “can cause code to run in a real system” as one permission forces an organisation to choose between a narrow, safe user base and a wide, risky one. Splitting them removes the choice: the population that may raise demand becomes very large, and the population that may cause code to execute stays very small and named.

It also makes external contribution commercially separable from external authorship — a customer can open the register to their own clients without opening anything else.

Entitlement-scoped workspace

An author works in a conversational workspace against their own data permissions, never a service account. What they can see in a conversation is what they can see in the application, which is the only arrangement that keeps an entitlement model meaningful once a model is reading on someone’s behalf.