The Next Phase of AI in Design

By Andy Hatch

For the past several years, the promise of AI in design, like so many other applications, has been framed around a single idea: automation. Prompt the tool and get the finished interface, brand system, prototype, or campaign asset. 

It’s easy to see why that story travels, since it makes creative work sound instant and scalable. While various agentic or MCP-powered workflows are showing real promise, in practice, the gap between demo and production has been hard to close. What’s emerging now is more practical. For design, one of the biggest value drivers for AI is less about one-shot perfection and more about giving designers superpowers with fine-grain control over how work gets made.

This theme was echoed at the 2026 Figma Config conference. The announcements felt less important as a product-by-product rundown, and more like evidence of a larger shift, one we’ve been tracking at Work & Co across AI-enabled work. 

Value is coming from focused tools shaped around real workflows, helping skilled teams do more of what they already do well. In design, that points to more controlled asset generation, faster prototyping, and less toggling between platforms. More work can happen inside the systems where teams already design and deliver, which is closer to how craft actually happens. 

“Making something interesting was easy,” said Itay Schiff, Creative Director of Weave at Figma. “Making something usable, for real projects, was hard. Creative work isn’t about single prompting. It needs references, iterations, consistency, and feedback.” I appreciated this honest observation as part of a product launch keynote and the acknowledgment of what capable professionals already know.

The most promising AI capabilities are those that enable designers to steer the creative process as work takes shape, with more power and precision at each step. A few developing features from Figma reinforced that direction:

• Weave brings node-based workflows and more precise asset generation directly into the canvas, giving designers more control over AI-assisted output.
• Code Layers lets designers import a GitHub repo, generate code from their designs, and edit both side by side on the same canvas. That helps close the handoff gap without asking designers to become engineers. 
• Agent Skills lets teams codify and share prompting workflows, so the hard-won knowledge of one designer can become infrastructure for the whole team. 
• Figma Agent can generate bespoke plugins, a capability with real implications for what a design deliverable can be. If designers can create small, custom tools for specific workflows, a deliverable can become more than guidelines, assets, and documentation. For brand systems, that could mean giving a client a reusable way to generate new expressions from defined inputs, in a specific style, for recurring needs.

Taken together, these aren’t features that take work away from designers. They extend what designers can do, tighten the connection between what gets designed and what gets shipped, and help the expertise behind a design system keep working after delivery. 

Overall, the “prompt to design” dream isn’t dead, but the larger opportunity looks a lot like what Figma announced at Config. It’s AI that makes skilled designers faster, gives them finer control over their process, and enables their work to keep delivering value long after launch.

About the author
Andy Hatch is a Director of Product Management at Work & Co. He works at the intersection of design, technology, and product strategy, with a particular focus on how AI is changing the way we work and the products we build.