Does Authority Survive the Consequence? A Continuity Model for Agentic Payments from Runtime Admissibility to Institutional History.

A bank can now preserve strong evidence that an agentic payment was allowed when it crossed a runtime control. That evidence does not, by itself, explain the history that follows. The payment may execute, settle, alter balances, close an invoice, attract a challenge, produce a return, and force corrections across several institutions. This study follows one synthetic USD 250,000 commercial payment through that boundary. It separates a single-use operational permission from the historical warrant that explains why the payment was released. It then makes four contributions. First, it defines a typed, append-only directed multigraph for admission, commitment, execution, settlement, reliance, challenge, compensation, and correction. Second, it provides a minimum record profile and an implementable JSON skeleton. Third, it states three propositions: authority does not pass between institutions; compensation does not erase the action it offsets; and correction must preserve four institutional clocks. Fourth, it turns the model into ten control requirements, ten adversarial tests, and a mapping to ISO 20022, UCC Article 4A, Nacha, card-network, and internal records. The result is conditional. An authorization record can anchor the later history only when every consequential transition adds its own authority, provenance, timing, and typed relation to the original commitment. The original permission does not survive. Its evidentiary value must.