How Workday Builds Its AI Interface On The Fly

Enterprise software giant Workday is building the user interface for its new AI assistant on the fly using a decision model.

Enterprise software company Workday has made its AI assistant, Sana, the new default home screen for its suite of HR and finance products. According to reporting from The New Stack, the most significant part of this launch is not the AI itself, but how its user interface is built. Instead of creating a fixed, pre-designed interface, Workday is using a tool called Jev to generate the UI dynamically. This means the interface is assembled in real-time based on the user's context and the AI's needs. This architectural choice represents a major departure from traditional UI development, especially within the typically conservative world of enterprise software, and offers a glimpse into how applications may be built in an AI-first future.

This approach, often called model-driven or server-driven UI, fundamentally changes the relationship between the front-end and back-end. In a traditional setup, developers write specific code for every screen, button, and interaction. At Workday, the front-end team builds a library of reusable UI components, but a 'decision model' determines how and when they appear. This model, powered by TypeSafe's Jev, likely takes various inputs—such as the user's role, their recent activity, and the specific task the AI is trying to help with—and then outputs a set of instructions for which components to display. This separates the application's logic from its presentation, allowing the user experience to be altered and personalized without shipping new front-end code.

The concept of a backend system controlling the user interface is not entirely new. Tech giants like Airbnb and Spotify have long used server-driven UI to update their mobile apps on the fly, enabling them to run A/B tests and roll out features without a full app store release cycle. What makes Workday's implementation notable is its application to an AI assistant. The conversational and unpredictable nature of AI interactions makes a static UI impractical. An AI might need to ask a clarifying question, present a data table, or offer a form for approval. A dynamic UI generation system can construct the perfect interface for each step of that conversation, making the interaction feel seamless and intelligent.

For developers, architects, and technology leaders, Workday's strategy provides a powerful, real-world case study. It demonstrates that model-driven UI is a mature pattern ready for complex enterprise use cases. This architecture can lead to faster development cycles, as UI logic can be updated on the backend independently of front-end releases. The key takeaway is the importance of decoupling presentation from logic, a principle that becomes critical when building for AI. As more companies integrate conversational AI into their products, we can expect to see wider adoption of dynamic UI generation, creating a new market for tools like Jev and shifting the focus of front-end development from building pages to building robust, adaptable component systems.

This showcases a shift from traditional, statically-coded UIs to dynamic, model-driven interfaces, especially for AI-powered applications. For developers and architects, it presents a viable, large-scale example of how to build more flexible and responsive user experiences that can adapt to complex, context-aware interactions.

Adopting a dynamic UI generation model can significantly reduce development cycles and costs associated with creating and updating complex user interfaces. This agility allows companies to innovate faster, personalize user experiences at scale, and gain a competitive edge by delivering more responsive and intelligent software.