Abstract: This article examines the evolving responsibilities of designers in an era of explosive AI growth. AI is a horizontal technology affecting nearly all industries, and designers must position themselves as stewards of the “human in the loop” to balance technological capabilities with human needs. Three core principles can guide ethical AI design: First, AI features should solve specific user problems rather than being implemented for novelty or marketing purposes. Second, strategic friction can serve as a beneficial design element when deployed at consequential decision points, encouraging users to engage thoughtfully with AI-generated content. Third, robust user feedback mechanisms should be prioritized to ensure continuous improvement based on real-world usage. Generative AI should facilitate — never replace — human expertise to avoid the centralization of ideas and displacement of creativity. Drawing from industry examples, the article demonstrates that successful human-AI collaboration depends not on technological sophistication alone, but on thoughtful design that empowers users as active participants rather than passive consumers of AI outputs. Eight recommendations are provided to ensure that the three core principles discussed are incorporated into a product’s design.

Implications for research: The recommendation to compare a potential AI-driven design solution to a viable solution predating AI (Section 2) suggests the development of a conceptual framework for systematically assessing and justifying aspects of AI specification. Such a framework or method would need to be compact to integrate into R&D. The principle of thoughtful friction in AI product design (Section 3) is introduced in relation to frictive AI elements with granular typographic detail such as point size and location. Empirical research could determine which typographic and interface characteristics provide friction, to what degree different options do so, and what thresholds exist for users’ emotional and cognitive reactions. Further research could examine whether such frictive elements effectively reduce trust in intentionally erroneous AI-generated content. Finally, the recommendation to integrate feedback mechanisms into AI-driven products (Section 4) suggests research aimed at understanding the current state of AI feature feedback collection in popular products, both in terms of mechanism and frequency.