Permission as a Variable: Why the same robot pays off in one plant and fails in another

Abstract

Projects that install the same robot model report labor-productivity gains an order of magnitude apart. Studies of robot adoption stop at the firm or the plant, so the gap between two plants with the same robot sits below the level they measure. Engineering studies stop at the machine. The missing variable sits between them: what a plant allows a specific robot to do on its own, on a specific job. That permission is set after the purchase, by people in the plant. Two plants can set it differently for the same robot, and no survey records it. Whether it moves afterward, as products, shifts and software change, is something this paper proposes to measure. It introduces the Habitat Readiness Score, which reads four organizational conditions for one permitted action against the job's stated requirement, and reports a profile rather than a total. It places the instrument beside safety readiness, which assesses an organization's readiness to manage risk and does not measure one action's permission against a job requirement. An archive of 295 documented robot projects, compiled by Hayashi from public handbooks, shows that the spread inside a single robot model is wide and that the pages do not report the evidence that would explain it. A three-step study is set out. Nothing here is tested. The paper offers the instrument that would make the claim testable. Working paper, version 0.20. Prepared for CORA 2026, the Conference on Robots and Automation, ZBW Leibniz Information Centre for Economics, Kiel, 28 September 2026, Session 2, Diffusion and Productivity. Case data compiled by Professor Ryuichi Hayashi, Kobe Gakuin University, from the METI and JARA Robot Introduction Demonstration Project handbooks, 2016 to 2018, used with his permission. The author has a commercial interest in this area through startup activity he invests in and helps build. JEL: O33, O14, D24, L23.