Agent-to-Business: An Episode-Based Framework for Delegated Machine Action Across Digital and Embodied Systems

Working paper, version 0.1 (4 November 2026). Conceptual research and design analysis; not peer reviewed. This paper develops Agent-to-Business (A2B) as an episode-based framework for analyzing delegated machine action across digital and embodied systems. Valid delegation, successful execution, and complete records are evaluated properties rather than prerequisites for inclusion. The framework connects six analytical dimensions: goals, normative authority, practical control, actual effects, institutional recognition, and evidence. A reusable procedure examines mismatches between these dimensions and between individual episodes and whole tasks. Constructed digital and robotic cases, a controlled conceptual comparison of delivery modes, and sixteen boundary cases show how the framework handles physical effects without digital receipts, disputed completion, and locally permitted actions that jointly violate a task. These constructions establish analytical consistency, not improved safety or user performance. The paper specifies comparative evaluations that could support, narrow, or reject its proposed analytical contribution. The longer-term concern is how machine participation redistributes decision rights, intervention capacity, consequences, and obligations to explain and remedy actions. Author: Jingchuan Nie, independent researcher and contributor to ACC and BailingHub. The paper discloses substantive AI assistance and the author's relationship to these projects. It is not an ACC normative revision or a report of new production or robot experiments.