The Orchestrator Standard: Professional Competence and Accountability in Directing AI
Year: 2026
Type: article
Abstract: AI oversight depends partly on the competence of the people assigned to it. The EU AI Act identifies competence, training, authority and support as conditions for human oversight of high-risk systems. Research on automation bias, misplaced responsibility and ineffective oversight gives reason to test those conditions rather than assume that appointing a person is enough. This paper proposes the Orchestrator Standard: seven competencies assessed through real work and an observed practical examination. Each competency has observable criteria and a defined disqualifying failure. Certification would require satisfactory evidence in every competency; strengths would not compensate for weaknesses. A candidate would have to produce an independently verified result or justify a decision not to proceed, while responding proportionately to faults and benign anomalies. The paper also proposes an independent Certification Board, routes for confidential evidence, continuing assessment and a small pilot with published progression and stop criteria. The Institute would develop the Standard; the Board would decide certification; the Orchestrator Awards would recognize exceptional work separately. The proposal has no demonstrated predictive validity. Its immediate purpose is to invite scrutiny of whether the competence can be assessed consistently, whether an additional credential is useful, and whether independent administration is feasible. Version 0.3, consultation draft. Manuscript dated 18 November 2026. Not peer-reviewed; not an adopted standard. Reading version: https://www.oiail.org/publications/the-orchestrator-standard
Author: Sendo, Mark
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236645928
