Design for AI: 6 Dimensions of Control in Agentic AI (+Prompt Inside)

A practical framework for user control in agentic AI, where it can get lost, and a prompt to evaluate your AI product or project.

AI agents and agentic workflows are now part of how we work and our everyday lives. An agent suggests next steps, runs workflows, makes decisions, remembers preferences, and sometimes acts across tools on your behalf.

The thing is, it does more than you can visually see, it takes actions you didn’t know about, and it might behave differently the next day on the same task.

I started to have questions about the feeling of being in control and what control is as an interaction pattern for AI systems. Questions about all the elements that allow people to get the benefit of automation without losing the feeling of ownership.

Here’s what's inside this article:

1. What control actually means in agentic UX

2. Six dimensions of meaningful control

3. The common patterns where control breaks

4. + a prompt to audit your own product

---

---