What AI Really Changes in UX Design NOW

What AI Really Changes in UX Design — NOW

High-level takeaway

AI is already useful in UX for design ideation, rapid prototyping, iteration and documentation, but is less disruptive for early-stage problem definition and primary research (user interviews/observations). Best use: accelerate repetitive, time-consuming “pixel-pushing” so designers can spend more time on research, strategy and creative direction. Biggest risks are loss of nuance, accessibility misses, generic outputs and baked-in bias from training data.

AI speeds tactical parts of the UX workflow (ideation → prototype) but doesn’t replace user research, UX thinking or creative direction.

Tools and workflows discussed

• ChatGPT: generate robust design briefs and produce JSON-style structured prompts to feed other tools.
• Figma Make (Figma AI): convert prompts/artboards/libraries into interactive prototypes; can create interactive text fields without coding.
• Figma First Pass / Google Stitch: generate scrollable static screens or static collections for different fidelity needs.
• Wave (accessibility) extension + BrowserStack: audit accessibility; Wave flagged many issues missed by AI-generated designs.
• Gemini and other LLMs: referenced briefly as additional options.

Workflow note: a useful prompt structure includes product info, product type, target users, product goals, user scenarios, main features and desired outcomes. Be specific initially; use concise, bite-sized prompts for iterations.

Concrete use cases shown

• From idea to prototype: ChatGPT → JSON prompt → paste into Figma Make → interactive prototype in minutes (example: an app for dog owners to date and check dog compatibility).
• Reskinning / branding experiments: attach artboards or connect a single Figma library to an existing prototype to rapidly test new look-and-feel directions. (Limitation: only one library can be connected to a prototype; workaround is to duplicate the prototype.)
• Iteration cycles: AI can quickly reskin and identify component styles, enabling fast directional experimentation across flows.

Strengths observed

• Speed: moves projects from 0 → 1 very quickly; accelerates directional exploration and repetitive visual changes.
• Prototyping fidelity: interactive text inputs and clickable prototypes without developer help are game changers.
• Component extraction: can identify and apply component styles across screens for rapid reskins.

Limitations and failures

• Interaction detail: AI-generated interactions can miss realistic gestures, nuanced animations and micro-interactions; often defaults to obvious controls (buttons).
• Accessibility: even when prompted to meet AA standards, AI often misses contrast, semantic structure and other issues — human QA and accessibility tools (e.g., Wave) are required.
• Visual nuance and polish: outputs can be generic and lack brand warmth or sophisticated visual treatments (nesting, layered backgrounds, iconography).
• Conflicting or redundant UX elements: AI may insert duplicate controls or elements that conflict with intended flows.
• One-tool constraints: outputs vary by tool (static vs interactive); choose tools to match the deliverable and expected client feedback.
• Ethical and bias concerns: fast adoption can entrench biases; designers must use tools mindfully.

Practical recommendations (tips & tactics)

• Use a structured prompt template: product info, users, goals, scenarios, features, outcomes.
• Be specific on first-pass prompts; use short, incremental prompts for iterative changes.
• Combine tools: e.g., ChatGPT for briefs, Figma Make for interactive prototypes, Wave for accessibility audits.
• Always run human QA for accessibility, usability, missing features and brand nuance.
• Treat AI output as a starting point — expect to refine for polish, interaction quality and edge cases.
• Use AI to free time for deeper research, strategy, experimentation and higher-differentiation creative work.

How AI changes UX practice (analysis)

• The practice doesn’t disappear: UX thinking, user research, creative direction and strategic work remain critical.
• AI will widen the gap between average (faster, more generic) and exceptional (deeper, more thoughtful) work — designers who master the tools and guard for nuance will have an advantage.
• Opportunity: quicker experimentation could enable more “weird” or playful brand-driven experiences at lower cost, if organizations allow exploratory risk.

Ratings / impressions

• Figma Make: strongly endorsed for ideation and prototyping (high marks for accelerating work), but outputs require follow-up for accessibility, polish and interaction fidelity.

Main speakers / sources

• Natalie — UX lead / design lead at Lefield Labs (LFL), primary presenter and tester of tools.
• Yan — interviewer/moderator (asked questions and probed implications).