Proposals

A proposal is a recommendation from an advisor that a human signs off on before it’s acted on. It’s the same as asking an advisor a question — but the answer is a decision that needs approval. Proposals are an organization feature (found under Enterprise → Proposals).

The maker-checker flow

1. Request — anyone picks an advisor and describes the decision they need help with.
2. Draft — the advisor (the maker) writes a recommendation, grounded in its datasources.
3. Decide — a checker or admin reviews it and chooses one of:
   • Approve — accept the recommendation as‑is.
   • Suggest changes — describe, in plain language, what to change; the advisor regenerates the full recommendation. Repeat until it’s right, then approve.
   • Reject — decline it, with a reason.

Every recommendation shows its evidence, and the final decision records who decided, when, and why — so the whole thing is auditable.

Note

You don’t edit the recommendation’s wording by hand. Instead you tell the advisor what to change and it rewrites the answer — so the result stays grounded in your sources rather than becoming free‑hand text.

Who can do what

• Anyone in the organization can request a proposal — the AI is the maker.
• Only a checker or admin can decide one. The person who decides is, by design, not the AI that drafted it — separation of duties is built in.
• Decisions are final: a proposal is decided once.