UX Collective: AI systems are demanding new interaction models. Are designers ready?
Arin Bhowmick, Chief Design Officer at SAP, published this five-minute article in UX Collective arguing that AI systems represent a qualitative break from everything designers have built mental models around. Where traditional software follows a turn-based pattern — user acts, system responds — AI systems enable continuous, parallel interaction where the work begins after a user states their intent. Bhowmick’s position is that designers have not fully reckoned with what this shift requires from them in practice.
What the article covers
The article builds around four categories of design challenge that become acute in AI-driven products.
The first is mental models. When an AI system interprets intent rather than responding to explicit commands, its reasoning is invisible to users. Bhowmick argues that designers must make that thinking visible — not because users want to see every inference, but because understanding what the system is doing is what allows them to correct it when it goes wrong.
The second is navigation architecture. Traditional interfaces have clear entry points, menus, and back paths. Generative and agentic interfaces are more fluid, which means discovery, orientation, and recovery from errors all require rethinking. The familiar spatial metaphors of screens and flows do not map cleanly onto systems that produce different outputs each time.
The third is intervention points. Rather than placing user control at every click, designers in AI-driven products need to decide strategically where to insert review, confirmation, and correction moments. Too many checkpoints slow the system down to the point of negating its value; too few leave users unable to catch errors before they propagate.
The fourth is human accountability. AI systems have contextual blind spots, and in many enterprise and regulated contexts, a human remains responsible for the outcome. Bhowmick writes that “where humans are accountable for the outcome, the UI is the instrument of that accountability.” The interface is not merely a wrapper around the AI; it is the mechanism through which responsibility is exercised.
What it does not cover
The article does not address specific tooling, design systems, or implementation patterns. It operates at the level of principles rather than methods, and it draws primarily from an enterprise perspective. Teams building consumer AI products in less regulated contexts may find the accountability framing applies differently.
Who it is useful for
Useful for product designers and design leads who are building AI-powered products and want a clear conceptual framework for what has changed about their work. Also relevant for design managers who need to articulate why AI product design is a distinct discipline, not a variant of the work designers already do.