Exploring the New AI Product Workflow
An experiment in AI-native product design and Design Systems
2026
The question
I've spent the last several years designing and leading Design Systems that help teams create digital products and experiences at scale.AI is starting to change how those teams work.Rather than simply using AI to generate content or ideas, I wanted to explore a bigger question:Can AI help carry a product idea from UX through to a working, validated experience?And if it can, what does that mean for the Design System?So I decided to build a small product from brief to prototype and document what happens.
The experiment
I'll take a realistic product brief through:Brief → UX → Design System → Prototype → QAusing AI alongside Claude, Figma and an AI coding environment.The aim isn't to prove that AI can replace designers or developers.It's to understand:What becomes faster?
What gets better?
What gets worse?
Where is human judgement still essential?
01 - Understanding the problem
I'll start with an intentionally imperfect product brief.AI will help turn it into a clearer product problem by identifying:- User needs
- Business objectives
- Requirements
- Assumptions
- Gaps and unanswered questionsI'll then challenge and refine the output.Can AI help move from an ambiguous brief to a clear product problem faster?The important part is not whether AI produces a good answer.It's seeing where it needs human judgement to get there.
02 - Exploring the UX
I'll take the refined problem into the UX process.AI will help explore:User journeys → Flows → Content → Key interactionsI'll test different approaches and use AI to challenge assumptions, identify edge cases and explore alternatives.The goal isn't to ask AI to “design the product”.It's to see whether it can act as a UX thinking partner.Can AI help explore more possibilities without slowing the process down?I'll compare the AI-generated thinking with my own decisions and document where the human contribution changes the outcome.
03 — Designing with a Design System
This is where the experiment connects most directly to my previous work.AI can already generate attractive interfaces.The more interesting question is:Can AI design within an existing Design System?I'll give AI access to the system's components, patterns, tokens and usage guidance.Then I'll test whether it can make sensible decisions about:- Components
- Patterns
- States
- Content
- AccessibilityIf it chooses the wrong component or misunderstands a pattern, I'll investigate why.Perhaps the AI needs better instructions.Or perhaps:The Design System itself isn't structured clearly enough for AI to use.
04 - From design to prototype
The next step is to move from Figma into an AI coding environment.The question:Can AI turn the experience into a working prototype without the traditional handoff between designer and developer?I'll compare the same journey using two approaches.TraditionalBrief → UX → Figma → Handoff → Development → QAAI-assistedBrief → AI-assisted UX → Design System + Figma → AI prototype → Human reviewI'll look at time, quality and rework rather than simply whether the prototype works.
05 - Testing the result
Finally, I'll use AI to review the prototype against:The original brief
UX requirements
Design System
Accessibility
User journeysI'll look for:- Missing states
- Incorrect components
- Accessibility issues
- Inconsistent behaviour
- Broken journeys
- Requirements that were lost along the wayCan AI become another layer of QA before human validation?
What I'm testing
01 - UXCan AI help explore and refine product experiences faster?02 - DesignCan AI turn UX thinking into useful interface solutions?03 - Design SystemsCan AI work within an existing Design System rather than generating generic UI?04 - PrototypingCan AI reduce the traditional gap between Design and Engineering?05 - QACan AI identify problems before human testing?I don't expect all five to work.That's the point of the experiment.
The failures matter
I'll document where things go wrong.AI misunderstanding the brief.AI choosing the wrong component.Good-looking UI solving the wrong problem.A Design System pattern being difficult for AI to interpret.A prototype that works technically but doesn't meet the original intent.These failures are important.They show where AI genuinely adds value, where it creates more work, and where human expertise remains essential.
What I'm trying to understand
The goal isn't to answer:Is AI good for design?It's to understand:What changes when AI becomes part of the product workflow?Does it reduce handoffs?Does it give designers more time for higher-value decisions?Does it change the relationship between Design and Engineering?And perhaps most importantly:What does a Design System need to become when both humans and AI are using it?
What I learned
[To be completed as I run the experiment]I'll document:What AI made faster
Where it improved the work
Where it created more work
Where human judgement remained essential
What this could mean for a real product teamThe experiment isn't about proving AI works.It's about finding out where it works - and what needs to change when it doesn't.