Design's New Baseline
AI is rapidly turning everyone into a potential product builder. As the ability to generate applications from a simple prompt becomes commonplace, design is poised to become the key differentiator. The question is what "good" means when the tooling is changing as fast as the output it produces. The emerging answer, drawn from the experience of building AI-powered features, is to look backward to the fundamentals of the craft, using AI to strengthen core design principles rather than replace them.
At Figma, designers working on its AI features have had to confront this challenge directly. Their focus has been on codifying the principles that make designs work, using AI to tighten the loop between design and code, and embracing pragmatic, incremental improvements. In each case, the goal is to use the technology to enhance human creativity rather than automate it into a generic template.
Codifying the Non-Negotiables
When teaching an AI to generate a first draft of a UI that uses a design system, the immediate challenge is teaching the mechanical rules of good design. LLMs are well-suited to text-oriented tasks like writing and coding, but generating good visual work requires a different approach. It was quickly determined that feeding an AI a massive, comprehensive rulebook was impossible—not only because defining every nuance of quality is impossible, but also because token limits make huge prompts technically infeasible.
The solution is the same approach used in teaching design. Deconstructing an intuitive understanding of quality into a few clear, concrete, and instructive rules that hold generally true for any interface is key. Effective rules are specific enough to apply in practice (such as "always place the primary action at the bottom of a screen" or "always measure in 8px increments") but universal enough to apply to almost any UI. By communicating these highest-leverage truths to the AI, the quality of generated outputs improved significantly within just a few weeks.
There is a wider point here. Even though software varies greatly, a surprising amount of it boils down to only a few core layouts. While this has caused some concern about homogeneity in the design industry, established patterns make technology easier for users to navigate. By mastering those foundational compositions first, an AI tool can establish a solid baseline. This consistent baseline becomes the new creative starting line, leaving the task of bringing genuinely new ideas to the table, and inventing the future, to the designer.
A Blurring of Disciplines
As AI begins to support both designers and engineers, the skills required for each role will begin to evolve. Designers may need more technical knowledge to command AI tools, while engineers may need a better grasp of design principles to build user-centric solutions. Both backgrounds are moving toward a common archetype of a "product builder"—a creative, multifaceted individual who can leverage AI to execute a vision.
This resembles the role of the design engineer, a profile that typically began as either a designer who learned to build or an engineer who refined UI skills. However, specialization will remain essential. Even as daily workflows blur, designers will continue to craft exceptional, nuanced user experiences, while engineers focus on building the underlying robust and scalable systems. AI may assist by offering layout suggestions or code snippets, but the final creative direction and technical architecture will remain human-led, resulting in a tighter, more seamless loop between design and development.
Start With the Mundane
A different perspective on good design involves focusing on pragmatic, high-impact solutions that work well today. Search is a clear example. Searching through files, tracking down teammates for links, and hunting for components or assets consumes significant time and effort. AI-assisted search changes this dynamic by allowing you to retrieve exactly what you are looking for simply by describing it, or even by uploading a screenshot. AI tools also breathe new life into the "design graveyard"—the valuable explorations that often fail to reach production. With AI-assisted retrieval, that past work can serve as a constant source of inspiration and help unblock creative stagnation.
This effectiveness depends on integrating AI into existing processes in a thoughtful, incremental manner. The benefits come from small, reliable features such as automated layer renaming or improved search. Waiting for an advanced, all-encompassing "AI assistant" risks overreaching and potential disappointment. By starting with what reliably works today, teams can let the AI's role expand in lockstep with its capabilities, supporting designers in going deeper on polish and delivering well-informed experiences without the distraction of chasing hyped—but unproven—visions.
Proactive Collaboration
Beyond incremental improvements, there is the model of co-creation. The ultimate utility of AI lies in its awareness of the user's specific context. Instead of tools that rely on constant prompting and risk producing context-free outputs, the most helpful AI will proactively understand your design systems, past work, and project goals. In this mode, AI collaborates and enhances by improving things as you work, similar to how GitHub's Copilot assists developers. This is less about an AI doing the work for you, and more about the tool working with you.
This requires the AI to index prior work. For example, after a team wraps up a planning session in FigJam, an AI with awareness of the board could identify meeting notes, priorities, and action items. From there, it could intelligently suggest creating a roadmap or a presentation deck that is infused with the team's specific visual language and structure. It does not necessarily generate generic, templated output into existence. Instead, it extends your team's existing way of working as a true creative partner.
