AI and the Design Process

How Figma Make and Figma MCP Server work together

Over the past few weeks, Figma has rolled out the beta of Figma Make to its pro-tier and higher users. More recently, it rolled out Figma MPC Server (also in beta). Both support the design and development process using generative AI, but in different ways.

Figma Make generates UI components, screen layouts, and interaction flows directly within Figma using text prompts (e.g., “create a mobile onboarding carousel with progress indicator”). The primary audience for Figma Make is designers and non-designers (PMs, etc).

Figma MCP Server enables AI coding tools (VS Code, Cursor, Claude Code, Windsurf) to fetch detailed design data (component details, styles, and layout data) directly from Figma. Figma MCP Server runs in Figma's Dev Mode and helps developers build UIs that more accurately reflect design specifications. (Not sure what an MCP Server is? I love this 2-minute introduction.)

You can imagine how these tools can work together in a design workflow: prototype in Figma Make -> finalize designs in Figma Design -> use Figma MCP Server to implement the design in code more faithfully.

8 essential tips for using Figma Make
☕ Medium Read (10 minutes)

Introducing our Dev Mode MCP server
☕ Medium Read (7 minutes)

Frontier Models

Do reasoning models really "reason"?

Another week, another updated reasoning model: in this case, ChatGPT o3-pro, which, thanks to OpenAI's dreadfully confusing naming, is more powerful than ChatGPT 4o. Reasoning models were also in the news with Apple's release of a controversial paper on reasoning models.

It's a good time to step back and understand how reasoning models work, and Ars Technica takes us through it:

"It's easy for laypeople to be thrown off by the anthropomorphic claims of "reasoning" in AI models. In this case, as with the borrowed anthropomorphic term "hallucinations," "reasoning" has become a term of art in the AI industry that basically means "devoting more compute time to solving a problem." It does not necessarily mean the AI models systematically apply logic or possess the ability to construct solutions to truly novel problems. This is why Ars Technica continues to use the term "simulated reasoning" (SR) to describe these models. They are simulating a human-style reasoning process that does not necessarily produce the same results as human reasoning when faced with novel challenges." - Benji Edwards writing for Ars Technica

Let’s talk about what AI “reasoning” actually does
☕ Medium Read (9 minutes)

AI in the Organization

It’s time to document like your AI depends on it — because it does

Even while understanding that most knowledge sharing in organizations occurs via social means, I've long been an advocate of organizational knowledge management (the systematic process of capturing, organizing, and sharing institutional knowledge). It requires time and resources within an organization (which many are reluctant to devote), but the payoff is a central source of truth that guides how work gets done.

In recent years, this type of organizational documentation has somewhat fallen out of favor, under the flag of keeping processes lean. However, in the age of AI, documentation becomes a strategic asset:

"The rise of AI-powered knowledge systems has transformed documentation from a cost center into a strategic asset. Well-structured documentation now serves as training data for AI systems that can multiply organizational productivity across all business functions.

...The “garbage in, garbage out” principle applies directly to organizational AI systems. Poor, inconsistent, or outdated documentation results in unreliable AI assistance across all business functions." - Frank Schillinger

The AI documentation paradox
🔍 Long Read (21 minutes)

Product Announcement

Sky reimagines AI-powered workflows on the Mac

Apple's native AI efforts have struggled rather publicly: AI-related functionality announced at last year's WWDC, including a vastly improved Siri, was nowhere in sight for this year's WWDC, and Apple still declines to give any dates on availability.

Meanwhile, the team behind Apple Shortcuts, who left Apple a couple of years ago, has previewed a very compelling, integrated AI experience for macOS: Sky.

What's interesting about Sky is not only its user experience (the demo video is definitely worth watching), but also that they built the application using only the APIs Apple makes available to its third-party developers. Sky is not yet available to the public, so its performance remains to be seen. However, the approach shows promise for how AI could integrate into our day-to-day computing lives.

Hi Sky (Demo Video)
Watch Time: 2 minutes

Sky Extends AI Integration and Automation to Your Entire Mac
🔍 Long Read (23 minutes)

AI in the Organization

What to do if you're in an organization that isn't embracing AI tools

Many designers are stuck waiting for their companies to "figure out AI" while competitors move ahead. Ben Blumenrose, Co-Director at Designer Fund, argues this is actually an opportunity: by learning AI tools on your own time, you position yourself to lead adoption when your organization is ready. His post offers practical strategies for building AI skills and demonstrating value to skeptical leadership.

LinkedIn Post
⚡ Quick Read (2 minutes)

That’s it for this week.

Thanks for reading, and see you next Wednesday with more curated AI/UX news and insights. 👋

All the best, Heidi