Skip to content
Article Smashing Magazine Sep 2026

Smashing Magazine: The death of the button — intent-driven interface design

The article by Carrie Webster, published on Smashing Magazine in September 2026, makes the case that the most forward-looking design goal is no longer to create more refined buttons or animations, but to eliminate the visible interface layer entirely. The argument rests on a shift from click-based micro-interactions to intent-driven experiences, where the user states a goal in natural language and the system executes it without requiring manual step-by-step navigation.

Webster illustrates the problem with a flight-booking scenario: a process that traditionally takes over ten clicks and several manual filters. Platforms like Perplexity demonstrate the alternative — a user states what they want, and the system synthesizes a response directly, removing the need to open multiple tabs or work through a rigid form flow. Vercel v0 does something similar for interface creation itself: a designer or developer describes a component in natural language and the system generates working UI code, bypassing the traditional design-to-handoff pipeline. Goblin.tools, a task-management tool, breaks down ambiguous goals into concrete checklists from a single input.

The design principles that emerge from these examples are grouped into three categories. First, high-level goal input: users express what they want, not how to get there. Second, generative UI: layouts and components are created dynamically in response to context rather than predefined. Third, autonomous execution: AI agents handle sequential steps in the background, retaining context across stages so that each step builds on the previous one rather than starting fresh.

Webster identifies two significant risks that designers must account for. The first is what she calls the illusion of control: when automation removes decision points, it can feel patronizing or unreliable, particularly for consequential actions. She recommends designing explicit confirmation prompts for high-stakes decisions rather than assuming full automation is desirable. The second risk is the black box problem: users who cannot diagnose what went wrong will lose trust in the system. The solution is visible, plain-language feedback that explains what the agent did and why, not just what the final output is.

For working designers, the practical shift is away from mapping static flows and refining individual components toward designing dynamic, adaptable UI patterns that respond to intent. Measuring success changes accordingly — the metric is not engagement or time-on-page, but whether the system solved the user’s problem and got out of the way. Teams building traditional button-and-form interfaces should read this as both a competitive signal and a prompt to reconsider what “a good interface” will mean in the next few years.