UX Collective: The interface has left the building

Om Prakash, a designer with seven years of experience building AI products including agentic retail systems at Walmart and a US patent on intent-based recommendation interfaces, opens this July 2026 essay with a pointed observation: the vocabulary designers have used for thirty years is breaking apart, splintering into multiple new dialects that barely resemble each other.

The central provocation is that agentic AI systems — those that don’t simply respond to a query but take sequences of actions, browse the web, write code, send emails, and fill in forms — have moved from research demo to production product between 2024 and 2025. This shift creates a fundamental design tension that Prakash frames as the core UX challenge of the current moment: how do you design an interface for an agent that doesn’t need an interface to do its work?

For most of digital design history, the interface was the product. The visual affordances, the navigation structure, the feedback loops — these were the substance of what a designer contributed. With agentic systems, the agent acts in the background. The user sets an intention and the system executes a chain of decisions autonomously. The traditional interaction design frame, where a designer anticipates each user gesture and maps it to a system response, no longer covers the territory.

Prakash describes three overlapping modalities that now coexist: chat-based interaction (text in, text out), voice interfaces where context shifts without a screen to anchor it, and agentic systems that run without continuous user input. Each has distinct first principles. Chat rewards density and precision in phrasing. Voice relies on conversational flow and disambiguation. Agents require a different kind of design entirely — one focused on goal specification, progress visibility, error recovery, and appropriate interruption points, rather than on the moment-to-moment choreography of button presses.

The piece does not offer a unified replacement framework. That restraint is part of its value. Rather than declaring a new design paradigm, Prakash maps the territory of what is actually changing: designers must now maintain fluency in all three modalities and understand where each one is appropriate. A product that works through ambient voice needs fundamentally different consideration than one built around a chat window or one that runs a multi-step agent in the background while the user does something else.

This is useful reading for designers working on AI-native products, as well as for product managers deciding which interaction model to build for. The argument is not that traditional UI skills are obsolete but that they are no longer sufficient on their own. The interface has not disappeared — it has multiplied and diverged.