The Work Behind Delegation: A Framework for Supervising AI Coding Agents

The Work Behind Delegation: A Framework for Supervising AI Coding Agents YEON SU PARK, School of Computing, KAIST, Republic of Korea NADIA ARVI, School of Computing, KAIST, Republic of Korea HAE RI LEE, School of Computing, KAIST, Republic of Korea SEHOON LIM, School of Computing, KAIST, Republic of Korea QIANOU MA, Human-Computer-Interaction Institute, Carnegie Mellon University, USA JUHO KIM, School of Computing, KAIST, Republic of Korea and SkillBench, USA Fig. 1. We introduce a new framework for supervising AI coding agents (bottom) by reconfiguring the five stages of Sheridan’s classical human supervisory control (top). Sheridan’s five stages—Plan, Teach, Monitor, Intervene, and Learn—map onto our seven stages—Plan, Monitor, Wait, Review, Teach, Manual Fix, and Update Assets—as shown by the connecting bands. As AI coding agents carry out development tasks with greater autonomy, developers are shifting from direct implementation toward supervising delegated work. Yet existing research offers limited understanding of how developers organize supervisory activities into connected workflows. Drawing on observations and workflow diagrams from 19 experienced developers, we reconfigure Sheridan’s framework of human supervisory control into seven stages and the connecting loops for supervising AI coding agents. We applied the framework to public developer discussions on Reddit and found that supervisory demands extend across stages and that developers manage them by concentrating effort in planning, delegating supervisory work to other agents, and turning recurring guidance into reusable assets. Our framework provides a useful analytical lens for understanding how developers supervise AI coding agents by capturing how supervision is structured in agentic software development.

Authors: Yeon Su Park, Nadia Arvi, Hae Ri Lee, Sehoon Lim, Qianou Ma, Juho Kim

Published in: arXiv

Publication date: 2026-09-21

Read the paper: https://doi.org/10.48550/arXiv.2609.24234

Source license: Creative Commons Attribution 4.0 International — https://creativecommons.org/licenses/by/4.0/

The authors and publisher do not sponsor or endorse this recording.