DEVOURED
Design isn't Dying. It's Shifting Left
Microsoft's design team argues that design work is moving from interface layers into AI model configuration itself, where the output is the experience.
What: Microsoft's Tech Futures team describes "shifting left" for design—moving from wireframes and UI patterns into the model layer, where designers must configure how LLMs behave for individual users based on their cognitive patterns, working styles, and context rather than creating predictable interface flows.
Why it matters: This reframes design's relevance in an era when many fear AI will eliminate design jobs—the work isn't disappearing, it's becoming more technical and moving deeper into the stack, requiring designers to shape behavioral patterns in probabilistic systems rather than static screens.
Takeaway: Start prototyping with real data and APIs instead of static design tools, and focus on defining behavioral patterns for different user types rather than one-size-fits-all interfaces when building AI-powered products.
Deep dive
- "Shifting left" traditionally means moving work earlier in the development process to catch problems before they become expensive—for design, it now means engaging with the technology stack (APIs, data, models) instead of only working in tools like Figma
- Microsoft's Tech Futures team designs in code and wires real APIs into prototypes to surface gaps that static screens can't reveal
- In model-forward systems powered by LLMs, the model's output is the experience itself, and probabilistic systems never produce the same output twice
- Design challenge shifts from predictable navigation patterns to behavioral consistency—a visual thinker should always get visual responses, detail-oriented users shouldn't get breezy summaries
- Different users have different cognitive patterns (granular vs big picture, analytical vs visual, concise vs elaborate) that should shape how the model responds to identical queries
- Personal working patterns vary dramatically and change over time, requiring systems that understand individual context like calendar patterns, work/life priorities, and communication preferences
- Human signals must be encoded at the model and intelligence level from the start, not bolted on later as interface features
- This requires understanding data behavior at its edges and what a "good" output means to specific individuals in specific moments, not population averages
- The core principle remains unchanged—design is about deeply understanding people and making systems work for them, just manifesting in different technical forms
Decoder
- Shifting left: Moving work earlier in the development process or deeper into the technology stack to identify problems before they become costly
- Model-forward: Systems where AI models like LLMs are the primary interface rather than traditional UI elements controlling the experience
- LLM: Large Language Model, AI systems that generate probabilistic outputs based on prompts and training data
- Tech stack: The layers of technology (databases, APIs, models, interfaces) that compose a software product
Original article
As AI reshapes product development, design is "shifting left" — moving deeper into the technology stack, now reaching into the model layer itself. In model-forward systems, the output is the experience, demanding behavioral design tailored to individual users rather than predictable interface patterns. The core principle remains unchanged: understanding people deeply so the systems built around them actually work for them.
Explore all 55 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi