Over the past few years, the Atomic design team has been defining our role on project teams—understanding where we add the most value and how we can better support other disciplines. We take a practical, hands-on approach to influencing and guiding the “human lens” of a software product. That means advocating for users, learning from them, and sharing those insights. That helps the team define solutions that people want to use, that solve real problems, can be built within constraints, and that stay on budget. In that time, we’ve added new tools and refined old ones. This evolution has coincided with the rise of generative AI. Over the past few months, we’ve been actively experimenting with integrating AI into our design work. These tools can help us move faster, reduce costs, and still deliver quality.
The Past Six Months: From Curiosity to Integration
We’ve been following developments in generative AI for quite a while, but it’s really in the past six to eight months that it’s increasingly become a core part of our design process. Small experiments have led to tools and practices that gained enough traction to stick around from project to project. Having worked on multiple projects using our adapted approach, I’ve come to appreciate the benefits—especially our ability to support a more diverse set of customers. Of course, there are trade-offs, but pricing pressures and clients’ desire for speed often push this direction.
The Power of a Centralized Source of Information
One of the biggest breakthroughs has been synthesizing project notes into a single, centralized source. This collective knowledge serves as a foundation for other tools and deliverables. In the past, we would hand off polished artifacts but leave a lot undocumented due to time and budget. With this approach, more of that valuable information is captured and ready to be built upon. Artifacts are more readily updated and referenced by team members and generative AI tools.
Prompting, Automation, and Process Alignment
Learning how to use prompting, meta-prompting, and automation tools to manipulate text, connect data sources, and streamline workflows has been a major asset. It’s allowed us to better align processes, tools, and best practices with Atomic’s approach—while supporting project-specific use cases. We have built out a number of tools that aid makers in parts of our process. The value is enough that I keep coming back to many of these tools, despite their imperfections.
Generative Production Design
Using the tools we’ve built and others available in the market, we’ve experimented with dynamically generating user flows, design systems, critiques, and wireframes—touching many parts of UX and UI production. With a deep understanding of the framing portion of projects, we’ve been asking how we can leverage these tools. Many times the results lack luster, but others have your jaw on the floor. These approaches still need guidance and oversight, but it’s easy to see the potential. The outputs of these tend to be imperfect drafts or inspiration that add a lot of value to a designer’s process. It will be interesting to see how someone with command of the process, access to these generative AI tools, and a combination of creativity and good taste can move remarkably quickly.
Prototyping in the AI Era
Prototyping has been an especially exciting area of exploration. With tools like Cursor or v0, designers can build components and clickable prototypes that allow their attention to detail to shine in the final product. On small teams, this can be a huge advantage, even though turning prototypes into production-ready code is still a bit elusive. Prototyping user flows or even entire applications takes a fraction of the time it took just a few months ago.
Looking Ahead
After six months of heavily testing and exploring integrating AI into our design work, it’s hard to imagine solely going back to the “old way.” Designers still need more tools and process support before the design workflow fully transforms, but it feels like we’re living through that change at warp speed. Challenges remain—ethical concerns, and managing the pace of change—but the value for customers has been impressive. They’ve been cautiously excited about the results so far, and I’m equally curious about what the next six to 12 months will bring: What new tools will emerge? How will they shape the role of design? How will they change how teams build software? One thing feels certain: we’re just getting started.