An agent can complete a task independently. That does not make the human irrelevant.
We are the fragile part of the interaction. Attention runs out. Fatigue weakens judgment. Understanding takes time. Systems should protect those limits, not overwhelm us with more output to review.
Design around humans means helping people reach informed, considered decisions. Surface the relevant context, explain uncertainty, make trade-offs visible, and leave room to think. The goal is not faster approval. It is better understanding of the problem and the consequences of acting.
This still matters when agents can do the work alone. Tasks have no value merely because they are completed. Their value comes from helping people.
Human involvement does not mean approving every automated step. It means shaping the purpose, setting boundaries, understanding consequential choices, and retaining the ability to intervene. It also means preserving opportunities to learn: sometimes working through a problem matters more than receiving the finished answer.
For developers building agent workflows, the question is not only, "Can the agent finish this?" It is, "Does this system help the person understand, decide, and grow?"
Automate the work without designing the human out of its purpose.
Based on Design Around Humans in the Turtleand Handbook.
Top comments (0)