Building Applications with AI Agents

Chapter 13. Human-Agent Collaboration

As agentic systems weave into our workflows, success depends as much on how they collaborate with humans on raw capabilities. This chapter brings together the interaction-level mechanics (interfaces, uncertainty signals, handoffs) and the governance structures (oversight, compliance, trust calibration) that turn opaque assistants into dependable teammates.

Effective collaboration depends on calibrating autonomy: knowing when an agent should act on its own, when it should ask a question, and when it should defer entirely to a person. We’ll walk through strategies for progressive delegation—starting with simple drafts or suggestions and building toward greater independence as trust grows—and we’ll highlight how to repair that trust if mistakes happen.

Roles and Autonomy

This section explains how agentic systems shift from human-guided execution to autonomous operation—and how human roles evolve to match. As agent systems gain autonomy, one of the most important questions becomes: what role should the human play? The answer is not static. It shifts based on the task, the stakes, and—most critically—the level of trust between human and agent. This section explores how those roles evolve over time, how organizations can design for progressive delegation, and what it takes to align people, processes, and expectations as agents become more capable collaborators. We begin by tracing the arc from executor to governor, then examine the organizational ...