A configuration in which several AI agents operate jointly on a task under human supervision, characterised by coordination among the agents and a human orchestration role; a structural property of the setup rather than a discrete event.
STATE. Coded per task configuration by (a) the count of distinct agents acting on the shared task and (b) presence of a human orchestration point that assigns or arbitrates among them (present/absent). A single-agent setup, or a multi-agent setup with no human orchestration point, is coded accordingly.
Proposed measurement protocol (not yet empirically validated): Configuration coding from system logs: number of concurrently active agents per task and a binary human-orchestration-present flag; the flag's inter-rater agreement could be reported as Cohen's kappa. Agent count is a direct log readout, not a rating.
https://andreasehstandlicenseofclarityloc.github.io/augmanitai-periodic/#topic-agentic-ai