Let the Agent Do It? How Software Practitioners Understand and Make Permission Decisions in Agentic AI Assistants

2026-10-05

AI Summary

This study addresses the challenges software practitioners face regarding opaque permission decision mechanisms and difficulties in understanding AI agent behaviors. Employing a sequential mixed-methods approach that combines in-depth interviews with large-scale surveys, this work conducts a comprehensive empirical analysis. It provides the first systematic investigation into practitioners' mental models of permission decision-making, introducing a novel perspective that distinguishes between "intent" and "authorization." Furthermore, it reveals how users dynamically adjust their supervision strategies based on observability, risk, and trust. To address deficiencies in existing permission systems, this paper proposes design recommendations centered on enhancing the visibility of action reversibility and optimizing approval logic, thereby improving the safety of human-AI collaboration.

Abstract

Agentic AI assistants increasingly act on developers'behalf by modifying files, executing commands, and accessing external resources. These actions often require permission, yet little is known about how practitioners make permission decisions while still benefiting from agent autonomy. To address this gap, we conducted a sequential mixed methods study, interviewing 18 practitioners who use AI agents and then surveying 115 practitioners based on the interview findings. We find that practitioners often understand agent behaviour through what they can directly observe and review, while decisions, data use, and other activity behind the scenes remain less clear. This uncertainty also shapes permission decisions, which depend on the scope and risk of an action, whether it fits the task, familiarity with the agent, and the environment in which it operates. Practitioners respond by adjusting how closely they oversee agents, from setting limits in advance to monitoring execution and reviewing work afterwards. How much scrutiny they apply depends on factors such as trust, task importance, time pressure, and the consequences of an action. Our findings suggest that permission systems should make consequential actions easier to review, distinguish what an agent is allowed to do from what the user intended, make reversibility clearer, avoid treating repeated approvals as stable preferences, and distinguish rejecting a single action from rejecting an entire approach.

Problem

• Agentic AI
• Permission decisions
• Software practitioners
• AI assistants
• Autonomy

Innovation

• Agentic AI
• Permission Systems
• Human-AI Interaction
• Mixed Methods Study
• Trust and Oversight