Design systems are a crucial component of modern digital product development, providing a unified set of guidelines and components to ensure consistency and efficiency. An often overlooked aspect of design systems is how they encode decision-making. While there is ample discussion about how to structure and organize components (e.g., versioning, atoms, libraries), I see less attention given to why certain decisions are codified at different levels of a UI component’s structure.
Consider a form field component. Are variations allowed at the label level, or the font size level? Both? What kinds of variations are or are not allowed? Can the form field only exist as part of a larger form, or can it stand alone? These are the kinds of questions that help determine which decisions are codified — that is, automated — and which decisions are left to the discretion of the individual designer employing the design system’s assets.
Automating certain decisions can be beneficial, enabling faster page creation and streamlining mockup or prototype creation. However, it can also hinder problem-solving when the constraints built into the system fail to account for unique or unforeseen situations. Achieving appropriate flexibility to account for a broad array of design requirements while maintaining brand and visual consistency requires careful consideration of which decisions to automate, which have a few standardized variant possibilities, and which must be left more open-ended.
These principles can also be applied to the implementation of AI-powered decision systems. Examining the flexibility required from the codified decisions is one of the ways to explore whether or not a particular decision may be ripe for automation. Decisions that require understanding novel situations, like those that fall outside a design system’s established conventions, are poor candidates for automation. Automating such decisions can lead to frustration and decreased productivity, particularly when the automation needs to be undone and the work redone manually.
Conversely, routine, low-risk decisions are likely suitable for delegation to an AI. In the design system analogy, these would be akin to font types, color tokens, and spacing — the smallest, most repetitive elements of the system. By automating these minute decisions, organizations can increase speed to delivery and decrease the frustration workers experience performing these tedious tasks.
Moreover, the delegation of decisions to an AI is ideally an iterative process, similar to the induction of new components (or the removal of deprecated ones) in a design system. As the AI system’s decisions play out, it’s crucial to continuously evaluate the net results. Establishing and maintaining this feedback loop enables refinements and adjustments to ensure that the AI’s decisions are the appropriate blend of flexible and predictable, and that they continue to meet the evolving demands of a dynamic business environment.
In many ways, a design system is a series of automated decisions. Reflecting on how and when conventions are codified or left flexible in a design system’s governance structure can provide valuable insights into an organization’s approach to automating decisions. By carefully considering the lessons learned from design systems, organizations can make informed choices about where and how to integrate AI into their decision-making processes, ultimately leading to more effective and efficient outcomes. If you’d like to learn more about how and where to integrate AI tools into your workflows effectively, schedule a consultation.