Is the App Layer Where AI Proves Its Value? | Figma Blog (opens in new tab)

AI’s next breakthrough may come less from larger models than from the application layer that makes them useful and accessible. Like graphical interfaces made personal computers mainstream, well-designed AI products can translate complex capabilities into intuitive, context-specific experiences. The products that succeed will combine reliable infrastructure with thoughtful interaction design and emotional resonance.

From MS-DOS to the App Layer

• Today’s prompt-driven AI resembles the MS-DOS era: powerful, but requiring users to know how to issue precise commands.
• Existing models have a “capabilities overhang,” meaning much of their potential remains difficult to access.
• Personal computers became mainstream through graphical user interfaces, not MS-DOS itself.
• Similarly, browsers, search engines, smartphone apps, and services such as Uber and Instagram transformed underlying technology into everyday tools.

Design Makes Technology Adoptable

• Building an app layer is not enough; adoption depends on the quality of the interactions surrounding the technology.
• Successful products combine functionality with intuitive design:
  • Pinch-to-zoom and inertial scrolling on smartphones
  • Live maps in Uber
  • Simple navigation in browsers and search engines
• AI products will need new interaction patterns that make model capabilities feel natural rather than like conversations with a raw chatbot.

AI Products Must Be Context-Specific

• Most people will use AI through specialized products rather than directly interacting with language models.
• Effective AI applications will adapt their content, tone, interface, and responses to particular audiences and situations.
• The Good Inside parenting app illustrates this approach:
  • It uses a chatbot trained on Dr. Becky’s parenting guidance.
  • Vague prompts receive empathetic, actionable advice.
  • Simple cards, a calm color palette, readable typography, and subtle animations create a reassuring experience.
• The same principle applies to products for lawyers, doctors, designers, artists, and other professional or consumer groups.

The Interface Can Matter More Than the Model

• User reactions to GPT-5’s simplified model picker showed that interface changes can provoke stronger responses than improvements to model capability.
• This does not make the underlying models unimportant, but users primarily experience AI through how its capabilities are packaged and presented.
• Atlassian’s acquisition of The Browser Company suggests that even browsers may evolve into active AI interfaces that help applications work together, rather than merely displaying tabs.

Design as a Competitive Advantage

• AI products will compete on the feelings and confidence they create:
  • Support for parents
  • Inspiration for artists
  • Confidence for lawyers
• Product teams must choose interactions that present AI outputs seamlessly while maintaining reliable, scalable systems.
• Many new AI applications will emerge, but the strongest may distinguish themselves through design and become as transformative as graphical user interfaces were for computing.

The practical opportunity for AI builders is to focus not only on model performance, but on designing specialized, emotionally resonant products that turn raw capability into useful everyday experiences.