From Co-Pilots to Co-Workers: A Formal Typology of Human–Agent Collaboration in Organizations

Abstract

Organizations are increasingly deploying agentic AI to function as semi-autonomous collaborators within end-to-end business processes. Yet, existing human–autonomy teaming research seldom addresses organizational roles, handoffs, and accountability in enterprise contexts. This paper introduces a design-oriented framework for human–agent collaboration that defines a typology of roles characterized by four measurable dimensions (autonomy, reversibility, criticality, and accountability) and a governance-oriented taxonomy of human control levels. The framework includes an indicative mapping to major AI governance standards (IEEE 7001 transparency levels, NIST AI RMF, and ISO/IEC 23894) and proposes process-level metrics such as task success, end-to-end latency, override rate, incident or near-miss frequency, and explainability coverage. Three enterprise scenarios (IT Operations, Customer Service, and Legal Operations) illustrate how to calibrate autonomy and oversight without eroding accountability, offering a practical guide for designing trustworthy hybrid organizations.

Keywords