Designing outcome-first “do it for me” products
AgentMost software helps you do things. Agentic software does things for you.
Instead of walking users through screens and options, agentic experiences let people state an intent: “book the cheapest train that arrives before 9am” or “draft a reply that matches my tone.” The system then plans and executes the steps across tools, forms and services, without requiring the user to manage each one.
That shift creates a new design challenge. Users aren’t navigating any more; they’re delegating. They’re handing responsibility to a system and trusting it to act on their behalf. Good agentic UX makes that delegation feel safe.
In practice, that means being clear about what the system is doing, what it’s about to do, what it needs from the user, and how to correct course if something goes wrong. It also means rethinking settings: rather than static preferences, controls become active guides for the system’s behaviour, covering things like budget limits, approval thresholds, preferred suppliers and rules around when to ask before acting.
When we design agentic experiences, three principles shape our approach: make the plan visible, require confirmation before irreversible steps, and always give users a straightforward way to undo. The goal is to remove the mystery from delegation, so users feel informed and in control, even when the system is doing the work.
