Outcome-Oriented Design
The shift from designing screens to designing outcomes isn't new thinking. The infrastructure to act on it finally caught up.
NNG recently published a piece naming "outcome-oriented design" as the next paradigm. Interfaces that respond to what a user is trying to accomplish, not just what they clicked. Design that starts with intent, not screens.
It's often called contextual UX, ephemeral UX, or adaptive UX. Different labels, same core claim: the interface assembles itself around what the user actually needs, right now. That's where everything is heading. Static screens built for average users and average contexts are already obsolete. Most products just don't know it yet.
This isn't an argument with NNG. Their framing is sharp and their reach is wide. The more people talking about this, the better. If naming it "outcome-oriented design" helps it land with resistant product teams, good.
But the line is worth drawing clearly - it matters for understanding what's actually new here.
The insight isn't recent. Designing for outcomes rather than screens isn't a new concept. It's been in the conversation since early service design work, since Jobs-to-be-Done got traction, since conversational UI pushed designers to think in flows rather than pages. What I've been watching, is that the AI infrastructure to act on this insight is now real.
What's new is that the model can now infer intent well enough to generate a surface in response. The gap between idea and implementation closed. That's the unlock.
So when NNG names "outcome-oriented design" now, they're right. They're also catching up to what practitioners have been watching at the edges for a while.
The design challenge that's still genuinely open: not just defining what outcomes we're designing for, but building the judgment layer that tells an AI system how to generate the right surface for a specific user in a specific moment. That work doesn't exist yet as a formal practice.
Outcome-oriented design is the what. The how (the frameworks, constraints, and principles specific enough for a model to act on) is the open problem. It's also the more interesting one.
If you want the longer thread on contextual and ephemeral UX. The thinking there hasn't changed. The infrastructure to support it finally caught up.