No. Every agent in the fleet proposes; a person decides.
An agent reads your registers, shows you exactly which rows and fields it would change, and writes only after an explicit yes. The change is recorded as agent-originated, so an auditor reading the register can tell what a person decided from what a machine suggested. Nothing is ever deleted; an entry that no longer applies is retired by status, because a register an auditor reads has to keep its own history.
This is a design constraint rather than a limitation we intend to remove. A register where a decision and a suggestion are indistinguishable is worth less as evidence, not more.