UX Archive UX Note
Human AI Interaction

AI experience design is decided
not by convenience,
but by when AI appears

Adding AI features has become much easier than before. What has become harder is deciding when AI should appear and when it should stay out of the way. Users are not always asking for AI. They want support only when it is needed, aligned with their intent, and explainable in the moment.

Section 01 — The Core Problem

Making AI more visible does not automatically make the experience better

In early AI product design, it is tempting to put AI-ness at the center. Chat, suggestions, summaries, generation, recommendations, and completion all look useful. But from a UX perspective, the question is not only whether the function exists. The real question is whether the user's intent and the strength of assistance match.

Too Much AI

AI that appears too much

  • Forces everything through chat
  • Suggests too much before the user acts
  • Over-explains every judgment
  • Hides the original work path
Right Touch

AI that appears with intent

  • Supports the moment confusion appears
  • Summarizes to reduce choices
  • Shows evidence for high-risk decisions
  • Lets the user regain control

The question in AI experience design is not "what can AI do?" It is how strongly AI should intervene in relation to the user's intent.

Section 02 — Placing Transparency

Transparency is not explaining everything. It is showing the necessary information at the necessary moment

With agentic AI, it becomes harder to see what the AI looked at, what it judged, and what it executed. But that does not mean every internal process should be exposed. Too much information becomes a burden, not transparency. What matters is identifying the moments when user anxiety or risk rises, and explaining the right thing there.

Moment 1
When AI chooses on behalf of the userBriefly show why that option was chosen and which conditions were prioritized.
Moment 2
When AI performs an action that affects the outside worldBefore sending, booking, purchasing, deleting, or sharing, let the user confirm the intent and result.
Moment 3
When AI is not confidentDo not assert through ambiguity. Show the conditions that need confirmation and the missing information.
Moment 4
When users look back laterImportant decisions, evidence, and changes need to remain searchable later, not buried only in chat logs.
Section 03 — The Limits of Chat UI

Chat UI is not universal. It was simply the easiest first form to ship

As an entry point for AI, chat is easy to understand. But in knowledge work and business processes, important decisions get buried inside conversation history. Chat alone is not enough when people need to find what was decided, why it was decided, and what should happen next.

Chat Fit

Where chat fits

  • Capturing intent from free input
  • Starting an ambiguous consultation
  • Clarifying conditions through dialogue
  • Answering short one-off questions
Structured Fit

Where structured UI fits

  • Preserving decision history
  • Comparing multiple options
  • Managing status
  • Letting teams see the same evidence

AI experience is not decided by whether chat sits at the center. It is decided by how units of work are shown. Conversation, cards, tables, timelines, diff views, and approval flows need to be combined according to intent.

Section 04 — The Designer's Role

The designer's role shifts from making screens to directing AI intervention

Generative AI has made wireframes, copy, screen ideas, and prototypes faster to create. That shifts human value toward deciding which option to choose, which risks to avoid, and which moments require AI to step back. In the AI era, UX design is differentiated not by the speed of screen production, but by the ability to design intervention.

Adding more AI does not create a better experience.
The AI that gets used appears only when needed
and quietly steps back when it is not needed.

The center of AI experience design is not the number of features, but the timing of intervention and the granularity of explanation.

References

Source Ideas Referenced in This Article

The article reorganizes and integrates the following source ideas as a practical view of AI experience design.

  1. The Rulebook for Designing AI Experiences
    UX Collective
    https://uxdesign.cc/the-rulebook-for-designing-ai-experiences-a22a50bb063c
  2. Identifying Necessary Transparency Moments in Agentic AI
    Smashing Magazine
    https://smashingmagazine.com/2026/04/identifying-necessary-transparency-moments-agentic-ai-part1/
  3. The Right Touch: Mapping AI Presence to User Intent
    UX Collective
    https://uxdesign.cc/the-right-touch-mapping-ai-presence-to-user-intent-d01fa2dee282
  4. Chatbox UI pitfalls / the conversation-forgetting problem in AI chat
    Integrated from collected CSV source rows