AI UX DAILY
Saturday, July 25, 2026
5 stories · curated for designers
The stories
Today in AI Products
 ChatGPT |
Jul 23 |
ChatGPT now connects to medical records and Apple Health for personalized health insights
OpenAI launched Health in ChatGPT for eligible U.S. users, letting them securely connect medical records and Apple Health data to receive personalized health insights through the chat interface. This is one of the first mainstream AI chat products to pull in genuinely sensitive personal health data and surface it conversationally. The feature raises immediate questions about how to communicate data scope, consent, and confidence levels inside a conversational UI.
| “ |
Audit how your product surfaces sensitive data in conversational interfaces — users connecting health records need clear, persistent signals about what data is being used, when, and what the AI can and cannot reliably conclude from it. — Designer's Takeaway |
| Webflow |
Jul 24 |
Webflow MCP 2.0 lets AI agents build from screenshots and edit design tokens in conversation
Webflow's MCP 2.0 upgrade lets AI agents generate components directly from screenshots, edit over 50 CSS-variable design tokens per conversation, and query analytics without a separate bridge app. A new 'Instructions' system lets teams embed always-on guardrails and on-demand workflows so agents follow team conventions automatically. This makes Webflow one of the first major design-to-production tools where a design system's constraints can be enforced through agent behavior, not just documentation.
| “ |
Consider writing explicit agent instructions in Webflow MCP 2.0 that encode your team's token naming conventions and component rules — this is the same design system governance work you'd do in Figma variables, now applied to the layer where AI actually builds. — Designer's Takeaway |
 Claude |
Jul 23 |
Claude voice mode expands to Opus and Sonnet and enters Gmail, Slack, and Canva
Anthropic expanded Claude's voice mode beyond the faster but less capable Haiku model, making Opus and Sonnet available for voice interactions. The rollout also extends voice access into third-party apps including Gmail, Slack, and Canva — meaning users can now hold extended, complex voice conversations with Claude inside the tools they already work in. Anthropic noted that early voice users quickly moved well beyond quick questions into longer, more demanding tasks.
| “ |
If your product integrates Claude or another voice-capable model, revisit your conversation flow design with longer sessions in mind — users are treating voice as a primary interface for complex work, not just a shortcut for quick lookups, which means error recovery and context handoff patterns matter more than you may have planned for. — Designer's Takeaway |
| Nielsen Norman Group |
Jul 24 |
NNG: As AI generates more UI, the designer's output is shifting from screens to curated context
NNG published a piece arguing that when AI generates interface elements, the designer's most important contribution is no longer the artifact itself but the UX knowledge fed into the generation process — user goals, constraints, edge cases, and design rationale. They call this 'UX-context design': structuring what you know about users so AI can act on it reliably. The piece is a direct response to the growing reality that a significant portion of interface work is being generated rather than hand-crafted.
| “ |
Start treating your research synthesis, user goals documentation, and design rationale as a first-class deliverable that gets maintained and handed to AI generation tools — not just as internal team artifacts — and audit whether your current documentation is specific enough to guide generated output. — Designer's Takeaway |
| Cognition / Poke |
Jul 24 |
Cognition acquires Poke to bring a distinct conversational personality to its coding agent Devin
Cognition bought Poke specifically to bring its conversational style and interaction model into Devin, their AI coding agent. The acquisition reflects a growing conviction in the industry that how an AI assistant communicates — its tone, pacing, and moment-to-moment interaction style — is a meaningful competitive differentiator, not just a polish layer applied after the core capability is built. TechCrunch framed it as AI personality becoming a genuine competitive advantage.
| “ |
Notice how product teams are now treating conversational personality as a core design artifact worth acquiring — if your AI product's interaction style was defined by defaults or engineering decisions rather than intentional design, this is a signal to revisit it before a competitor's tone becomes the standard users expect. — Designer's Takeaway |
Today's Idea
The interface layer is expanding, and designers need to own more of it
This week's releases make one thing clear: the surface area that designers are responsible for is growing faster than most teams have updated their process to handle. Voice modes are running complex sessions inside Slack and Canva. AI agents are building components from screenshots inside Webflow. ChatGPT is surfacing medical records in a chat window. In each case, the design decisions that matter most — what context AI gets, how it communicates uncertainty, what personality it carries — are being made with or without a designer in the room. The teams getting ahead of this are treating research documentation, agent instructions, and conversational tone as designed artifacts, not handoffs.
Stop shipping AI slop
Audit your AI design against 36 patterns
Drop a screenshot, get specific gaps and a Claude Code prompt to fix them. Free, no signup for the first audit.
Audit your design →