Seven Verbs: A Theory of Human Judgment in AI-Augmented Production Systems

Abstract

This paper proposes a seven-verb taxonomy for characterizing human work in AI-augmented en- vironments. Through dual independent deriva- tion—from organizational outputs backward and from irreducible human capacities forward—we identify seven verbs that describe the complete set of human contributions that persist when artificial intelligence handles execution: WANT (originate in- tent), INTERROGATE (pressure test output), TASTE (judge quality), OWN (bear consequences), ATTEST (stake reputation), WITNESS (provide presence), and CONVENE (align humans). We validate this taxonomy against 1,000 occupations in the U.S. De- partment of Labor’s O*NET database across 23 job families. All seven verbs load across all job families; none are orphaned. The distribution reveals ATTEST (85%) and TASTE (68%) as near-universal, INTER- ROGATE (47%) and OWN (44%) as core to decision roles, and WITNESS (27%), WANT (25%), and CON- VENE (19%) as specialized human-presence func- tions. We argue that the primary economic value of AI augmentation lies not in labor substitution but in decision quality improvement: same people, same hours, much better decisions. We further identify failure modes when each verb degrades and derive implications for organizational design, governance, and workforce development.