Workday has implemented TypeSafe’s Jev, an AI decision model, to generate user interfaces on the fly for its AI assistant Sana, enhancing speed and interactivity without hard-coding UI logic.
- Jev enables sub-second UI composition using atomic design components
- Decision-model approach reduces cloud compute costs compared to large generative models
- Integrates with developer workflows for responsive, real-time form pre-filling and input handling
Infrastructure signal
Workday’s adoption of TypeSafe’s Jev for generating user interfaces highlights a shift towards decision-model-driven UI construction rather than fully generative approaches. This introduces more predictable compute requirements with Jev answering interface element selections in 70 to 500 milliseconds, significantly faster and more cost-effective than traditional large language model-based generation. For cloud cost management, this means lower processing time and reduced reliance on heavy GPU or CPU resources.
The Jev model relies on atomic UI components drawn from Workday’s design system, ensuring consistent and reusable interface elements. This induced predictability allows more stable cloud resource planning and contributes to better reliability. Data preloading plays a crucial role in responsiveness, requiring infrastructure that supports efficient streaming and data availability to feed Jev’s decision logic and deliver near-instantaneous UI updates.
Developer impact
Developers benefit from Jev by simplifying the complexity of creating generative user interfaces. Rather than coding specific interface logic, developers supply Jev with all permissible design components and typed question sets. Jev returns selected interface elements, scores, or boolean answers with associated probabilities, reducing coding overhead and improving iteration speed for UI workflows.
The integration with Workday’s AI assistant Sana enables dynamic, context-aware UI generation such as forms auto-filled in real-time and interactive cards shown instantly as users type. This improves developer workflow by enabling a modular, data-driven approach where closed form options are neatly handled by Jev and fallback handled by language models when uncertainty is high, creating a more maintainable and extensible system.
What teams should watch
Teams focused on cloud infrastructure should monitor the cost and performance implications of decision-model-driven UI rendering compared to traditional generative AI frameworks. Given Jev’s speed and efficiency, real-time interface generation at scale becomes more achievable without extensive infrastructure overhead—this could influence platform decisions related to backend scaling and caching strategies.
On the developer side, UI/UX teams and frontend engineers should explore integrating similar decision models that leverage atomic design systems coupled with real-time data preloading. Observability tooling may need to evolve to track model confidence and fallback behavior, ensuring quality control and debugging in production UI environments. Additionally, product teams should assess how this approach impacts user experience consistency and customization.