Governing Delegation to Generative Artificial Intelligence: Human Direction, Work-Related Orientation, and Modes of Use

The delegation of cognitive operations to generative artificial intelligence (AI) presents a critical governance challenge, as human direction remains essential. This analysis identifies two primary modes of human direction: specified delegation, which pre-defines tasks through instructions and constraints, and iterative coproduction, which involves ongoing interventions during AI output generation. These modes are examined using data from the Anthropic Economic Index, distinguishing between direct API usage and conversational/collaborative AI platforms.

Why it matters

This research provides a foundational understanding of how human direction integrates with generative AI delegation, which is critical for establishing effective governance models. It highlights the need for organizations to strategically choose and implement appropriate modes of AI interaction, impacting operational efficiency, risk management, and the overall value derived from AI investments.

Key insights

• Delegating cognitive operations to generative AI necessitates a clear framework for human direction to address governance concerns.
• Two distinct approaches to human direction are identified: 'specified delegation' (pre-execution direction) and 'iterative coproduction' (during-execution interventions).
• The analysis uses the Anthropic Economic Index (April-May 2026) to assess these approaches.
• Data sources include 1P API (direct API traffic) and Claude.ai (combining Chat and Cowork activity) to differentiate modes of AI use.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.17624