A duo mapped 14 recurring ways people actually experience AI products — what might be on that list?

This explores the catalog of 14 experience patterns that Josh Clark and Veronika Kindred proposed for designing products built on AI. The collection doesn't contain that list, and listing the 14 from memory would risk inventing them. What it does have from Clark is a smaller framework. He describes three ways to organize multiple AI agents, each modeled on how people work together: pipelines that pass work from one agent to the next, critique loops where one agent refines another's output, and swarms where a conductor agent hands out parallel tasks What organizational patterns work best for multi-agent coordination?. The note adds a useful caveat. Swarms assume work can be split into clean, separate pieces, and research on how context gets fragmented across agents challenges that assumption.

If what drew you to the question was 'what are the standard design patterns for AI experiences?', the closest material is Victor Yocco's six UX patterns for AI agents. They're grouped by when they apply. Before the agent acts, there's a preview of what it intends to do and a dial for how much autonomy it gets. While it acts, it explains its reasoning and signals how confident it is. Afterward, there's an audit trail with undo, plus a way to hand the task back to a human Do six UX patterns improve agentic AI user trust?. Each pattern targets a specific psychological need, such as reducing surprise or making mistakes recoverable. That's a handy test for any pattern catalog, including Clark and Kindred's: what user anxiety does each pattern resolve?

The collection also has patterns that come from the opposite direction: they describe what users actually do, not what designers prescribe. In 400 real sessions with web-browsing agents, people settled into four stable styles: Hands-off, Hands-on, Collaborative, and Full Takeover Do users develop stable collaboration styles with web agents?. Separately, interviews found that users hold four incompatible beliefs about how ChatGPT's memory works, and those beliefs shape what they're willing to share What mental models do users hold about LLM memory?. Prescribed patterns like Yocco's work best when they match these real behaviors. An autonomy dial, for example, is really a way of letting each of those four collaboration styles coexist in one product.

To get the actual 14 patterns, go to Clark and Kindred's own writing. Their book *Sentient Design* is the likely source. The collection currently covers only Clark's multi-agent piece.