Designing With AI: A UX Process Guide

When it comes to AI for design, workflow is everything. AI is not one magic button. The real power is mapping AI to the right stage of design.

As designers we should be excited to add these new tools to our process. We must become proficient in not only using AI for admin work or generative brainstorming but grasp how to apply it uniquely at every stage of our design process. I understand the limitless possibilities with AI, I dated ChatGPT one time. Lets just say that it was an amazing imaginary evening! But I’ll reserve that story for another time.

As designers of the future we need to widen our scope and fully embrace how AI will make us better at what we do. We are not in a competition if we consider ourselves ever learners this should represent a new frontier.

I am writing this to start a conversation around how we can use AI in every stage of our workflow. To be clear, I am not trying to define anyone’s design process or tell you how you should work. what I am doing here is simply calling out the common stages that most design work moves through, whether you do them in order, overlap them, or loop back and forth. Think of this as a framework, not a rulebook.

The purpose is to show how we can harness AI at each stage of the workflow, from understanding the ask all the way through iteration. When we map AI to the right moments in the process, it becomes a tool that supports our work instead of distracting from it.

Lets break our workflow into 7 stages:

• Discovery / Research / Onboarding
• Framing Our Discovery
• Ideation
• Designing Assets
• Creating Design Systems
• Building + Validation
• Iteration

**Discovery / Research / Onboarding: **Understand the Ask (Intake + Problem Definition)

Activity Space: Good design starts before we can ever create a deliverable. This stage is about finding clarity from a lot of ambiguity: what is the ask, what are we solving, for who, and how will we measure success?

Best AI to Use: This is where we unleash AI to organize chaos. Feed it sows, notes, emails, requirements, stakeholder opinions. Ask it to summarize goals, assumptions, risks, constraints, and unanswered questions. It won’t replace conversations, but it will speed up alignment.

**Framing Our Discovery: **Frame the Experience (Journeys + IA)

Once the ask is clear, we frame the experience. This is where we map journeys, top tasks, and information architecture so design decisions have structure.

AI can draft journey steps, identify user intent, and propose IA models. It can also spot gaps and contradictions early. The designer still decides what is true and what is feasible.

**Ideation: **Concepts + UX Directions

Ideation should be fast and grounded. Not just “cool,” but “right.”

AI is the perfect brainstorm partner because it gives range. Multiple layout directions, flow variations, content patterns, and UX approaches in minutes. It also helps write clear concept rationales so stakeholders can make decisions faster.

**Designing Assets: **Wireframes → UI

This is where structure turns into screens. Wireframes become UI, responsive behavior, and detailed interaction design.

AI can speed up microcopy, empty states, error states, accessibility checks, and design variations. It helps you explore more options without getting stuck too early.

**Creating Design Systems: **Optimize for Dev Alignment

Design systems are how UX becomes scalable. If you want consistency and speed, you need tokens, components, and rules.

AI can help document components, define states/variants, suggest token naming, and generate guidelines. This is where design and development begin to feel like one system.

**Building + Validation: **Prototype / “Vibe coding”

This is where ideas become real. Many projects fail in the gap between design and build.

AI makes prototyping faster and more realistic. It can generate starter front-end code, translate UI logic into components, and troubleshoot. Designers who can “vibe code” will be extremely valuable. Companies will pay for someone who can take them from zero to hero in weeks versus months.

**Iterating: **Measure & Improve

Design is never done. It is launched, measured, and improved.

AI helps interpret analytics, summarize patterns, generate experiment ideas, and turn insights into a backlog. Iteration gets faster, and decisions get sharper.

Final Thoughts

Jensen Huang says we should become proficient in directing an AI, prompting an AI, establishing guardrails, and evaluating the AI. That last part is everything.

Prompting is easy. Evaluation is expertise.

AI will not replace designers who can think, decide, and build. AI replaces the parts of our workflow that were never creative in the first place. And honestly, good.

Appendix

Best AI Tools for Each Stage

Discovery / Research / Onboarding

• ChatGPT (summarizing, clarifying, and pulling out risks)
• Claude (for long documents like SOWs, meeting notes, and requirements)
• Notion AI (for project notes and tasks live in Notion)
• Microsoft Copilot (for Outlook, Word, Teams)
• Google Gemini (for Google Docs, Gmail, Google Drive)
• Otter.ai / Fireflies.ai (for turning meetings into summaries, action items, and key decisions)

Framing Our Discovery:

• ChatGPT (great for journey mapping drafts, task models, IA ideas, taxonomy options)
• Claude (excellent for pulling patterns from research and synthesizing themes into journeys)
• Miro AI (helpful for turning messy stickies into grouped themes and journey stages)
• FigJam AI (great for clustering notes, summarizing workshops, and mapping flows)
• Dovetail AI (best for UX research synthesis into themes and insights that fuel IA decisions)
• Optimal Workshop (AI-supported insights) (helpful when validating IA through card sorting / tree testing)
• Notion AI (good for organizing IA documentation and journey writeups)

**Ideation: **Concepts + UX Directions

• ChatGPT Best all-around brainstorm partner for generating UX concepts, flow variations, and structured pros/cons.
• Claude Excellent for deeper reasoning, concept framing, and writing strong rationales that sound human and strategic.
• Perplexity Great for quick idea generation with references (helpful when you want to ground concepts in real-world patterns and examples).
• Miro AI Helpful for turning messy brainstorms into structured themes, buckets, and concept clusters.
• FigJam AI Great for rapid workshop synthesis: sticky clustering, concept grouping, and idea expansion.
• Relume AI (web UX / sitemap ideation) Super useful for quickly generating sitemaps, page sections, and user flow structure for websites.
• Uizard Great for fast concepting from text prompts into UI drafts (especially early-stage).
• Midjourney / DALL·E Best for visual direction ideation: mood, style exploration, hero illustration concepts, vibe references.

**Designing Assets: **Wireframes → UI

• Figma AI / Figma Make Great for turning structure into screens quickly, generating layout variations, and speeding up production.
• ChatGPT Best for microcopy, CTA variations, empty states, error messages, and interaction behavior suggestions.
• Claude Excellent for longer UI copy, consistent tone of voice, and refining messaging across multiple screens.
• Magician (Figma plugin) Useful for quick UI content generation, icon ideas, and fast iteration inside Figma.
• Stark (Figma plugin) Best for accessibility, including contrast checks and visual impairment simulations.
• Maze AI Great for testing wireframes/prototypes and getting AI-supported insights on usability issues.
• Midjourney / DALL·E Strong for visual exploration like hero imagery, illustration concepts, and style direction references.

Creating Design Systems:

• Figma AI + Figma Variables Best for organizing tokens (color, type, spacing), scaling design system decisions, and speeding up component structure inside Figma.
• ChatGPT Great for defining token naming conventions, writing system rules, documenting components, and creating dev-friendly guidelines.
• Claude Excellent for long-form documentation, component usage guidelines, and keeping design system language consistent across many components.
• Zeroheight (with AI features) Best for design system documentation and publishing guidelines in a way dev and design teams actually use.
• Storybook + AI assistants (ChatGPT/Copilot) Best for bridging design to dev by generating and refining component code, writing stories, and improving documentation tied to real UI components.
• GitHub Copilot Great for supporting engineers (and designer-builders) in building components faster, staying consistent, and reducing implementation drift.
• Specify (design-to-dev automation) Strong for syncing tokens and assets from Figma to code and helping teams manage a source of truth.

Building + Validation:

• Replit AI Great for rapid prototyping in a browser-based environment (easy to build + test quickly without local setup).
• Vercel v0 One of the best tools for generating production-ready UI components and layouts (especially React + Tailwind).
• Locofy.ai Strong for converting Figma designs into front-end code and components (useful for fast prototyping and dev handoff).
• Anima Helpful for generating responsive, dev-friendly prototypes from Figma and exporting usable front-end code.
• Maze AI Best for validation testing, fast usability feedback, and AI-assisted insight summaries from prototype tests.

Iterating:

• Amplitude (AI / insights features) One of the best platforms for product analytics and behavioral insights, with AI-assisted pattern detection and segmentation.
• Mixpanel (Signal / AI features) Strong for funnel insights, retention, cohorts, and surfacing what drives conversion behavior.
• Hotjar AI Great for turning session recordings + heatmaps into summaries of behavior patterns and friction points.
• FullStory (AI-assisted insights) Excellent for identifying usability friction through real session replay data and surfacing pain patterns.
• Dovetail AI Best for synthesizing qualitative research and feedback into themes, insights, and opportunities that become backlog items.
• Maze AI Great for turning usability test results into actionable findings and identifying what to iterate next.
• Notion AI / Confluence AI Helpful for documenting insights, writing experiment plans, and maintaining an iteration backlog that stays clear and shareable.