The Future of AI: "Wigit" for computational design and prototyping

Microsoft

Jul 31, 2025

The Future of AI blog series is an evolving collection of posts from the AI Futures team in collaboration with subject matter experts across Microsoft. In this series, we explore tools and technologies that will drive the next generation of AI. Explore more at: Collections | Microsoft Learn

The Future of AI: "Wigit" for computational design and prototyping

AI isn’t just reshaping what we build—it’s transforming who gets to build. With today’s AI-powered tools, rapid prototyping is no longer limited to seasoned developers or designers; anyone can now bring ideas to life in minutes. These tools are quickly shifting from nice-to-have enhancements to essential parts of our daily work.

On the Azure AI Foundry Product Team, I’ve been building internal AI prototyping tools that let us instantly test ideas, while eliminating the usual trade-offs of time and cost when exploring what’s possible. I believe that using AI in this way can amplify everyone’s creative potential and accelerate innovation – all while democratizing the ability to do so.

One of the tools I've made for internal use, along with a team of computational designers, is called “Wigit.” It is a tool we created using AI to help us deliver product and communicate more efficiently. However, this case study is about more than just Wigit. It is about adopting a new mindset for creating custom tooling in the age of AI and amplifying everyone’s ability to effect change.

How We Prototype Software in Microsoft CoreAI: Wigit

Moving Beyond Static Artifacts

Traditional prototyping often follows a process where designers create static images, stakeholders give feedback, and eventually the images are passed to developers to be built into code. Each stage introduces the potential for delay and increases the risk of miscommunication. By the time a prototype becomes interactive, the context or requirements may have already shifted.

Having been a developer and designer for many years, I'm always looking for better ways to work and reduce friction in my day-to-day workflow. From my experience, prototypes are the most effective way to share ideas and to quickly sense whether an interaction truly fits the problem. Consider this: what if the prototype itself became the specification? What if anyone could describe a UI prototype in plain language and see it rendered immediately as interactive live code? What if further iteration could happen directly on working software from day one?

While many excellent tools like this exist, we chose to build our own version—using iterative discovery and Azure AI Foundry—to accelerate the journey from idea to outcome.

Wigit: A Case Study in AI-Driven Prototyping

Wigit started as an internal experiment inspired by my appreciation for REPL's (Read Eval Print Loop), which are great for testing small pieces of code. I envisioned a browser-based REPL for HTML, CSS, and JavaScript, enhanced with an AI chat interface at its core. The goal was to empower any team member to transform concepts into interactive prototypes within minutes, using web standards for maximum compatibility and portability. I also wanted to keep the code visible to users, believing that tools like Wigit can serve not only as powerful prototyping platforms but also as effective teaching resources.

How Wigit Works

Natural Language to Code: Users describe a data table, a search form, or even a simple game. Wigit’s AI generates and updates the code and UI in real time.
Live, Interactive Output: Every change is rendered nearly instantly. There are no build steps or delays.
Multi-modal Editing: Users can upload sketches, select UI elements directly, and provide specific instructions along with those assets.
Project Management: Wigit provides auto-versioning, easy import and export, and shareable URLs to support collaboration.
Plain Code Export: All output is standard HTML, CSS, and JavaScript. Prototypes are easy to port to production or integrate into frameworks. The prototypes are also stand-alone HTML pages that can run in any browser directly.

Why Build Instead of Buy

While there are many commercial tools that offer some similar features, few offer the flexibility and control we needed for our application. Additionally, some off-the-shelf solutions did not meet our needs with respect to privacy, extensibility, and long-term support. By building Wigit in-house, we achieved three objectives:

First, we met our enterprise security and privacy requirements by leveraging Azure AI Foundry.
Second, we gained full control over customization. With direct access to the codebase, we can evolve Wigit as our needs change and integrate it closely with our internal workflows.
Third, we were able to "ground" the model in a couple of our design systems, like Fluent, to help give a more realistic output for our needs.

AI as a Force Multiplier for Custom Tooling

Through developing Wigit, I saw firsthand how AI transforms the economics of creating custom internal solutions. In the past, creating a custom REPL or integrating natural language processing (NPL) would have required substantial expertise and time. Today, AI models have made these capabilities far more accessible, enabling teams to realistically build custom tools—something that was once only feasible for large software vendors. AI allows almost anyone to create these tools now.

What Wigit Enables

Wigit is bridging the gap between our design and implementation teams - helping product managers, designers, and developers to work together in a unified, interactive environment from the very beginning. Fully functional prototypes allow for early user testing, and when requirements shift, updates can be made in just seconds.

This approach has already improved and informed several internal projects. The workflow is expected to evolve from static assets to dynamic, living specifications. Developers can then spend less time interpreting intent and more time refining solutions. Feedback cycles can be shorter and more effective.

One of our designers recently used Wigit to create a working example of a feature's accessibility reflow across different screen sizes. This made it possible to provide a link to the exploration, enabling the developer to access both the code and the prototype. Without Wigit, this would typically require many screens of Figma designs as well as documentation about when the changes between the screens should occur, as well as back and forth with developers about what is and isn't possible. With this prototype, the designer was able to do all of that at once, in much less time.

Rethinking Internal Processes

My experience building Wigit suggests that teams can also reconsider the way they approach tool-building and innovation. AI makes it practical to create custom tools that fit specific workflows, security needs, and team culture. If your current process relies on static handoffs or slow feedback loops, consider how an AI-assisted environment could remove bottlenecks and increase clarity for your team. Features like natural language prompts, instant rendering, and multi-modal input can transform both workflow and output.

Looking Forward: AI-Built, AI-Driven

Wigit isn't just an internal tool; it is proof of concept for a new approach. We used Azure AI Foundry to create a tool that itself uses AI to help us build better products. As AI becomes more central to engineering and design, the boundary between tool and teammate will continue to blur. The tools we use will adapt to us as quickly as we adapt to them.

We are only beginning to explore what is possible when AI is at the heart of internal product development processes. The next wave of innovation will be defined not just by what AI can do for users, but by how it can help us all become builders and innovators of the future.