How I use AI in my design process
I use AI tools to support and speed up how I work.
The order of my design process hasn’t changed. The tools have. This is how I’m using them right now on Rayo at Bauer Media: what’s working, and what I still do myself.
My AI journey, in 6 months
01
Very sceptical
Only used ChatGPT now and then.
Worried about racial bias in what it generated, and I still keep an eye on that.
02
Trying things
ChatGPT and Miro AI, mostly for talking through ideas and summarising.
03
Using it for real
Figma Make and Figma Agents, on actual design epics.
04
Connecting it up
Claude with our real components, for rapid prototyping and user testing.
My design process
Whether it’s a new feature, a customer problem or improving something already live, my process is the same:
1. Kick-off
2. Research & Discovery
3. Ideation & Exploration
4. Prototyping & Testing
5. Iteration
6. Delivery
AI shows up at different points, mainly in four areas:
01
Understanding the problem · Kick-off + Research & Discovery
02
Ideation & exploration
03
User testing & production · Prototyping & Testing + Iteration
04
Presentations · Delivery
01 / 04
Understanding the problem
Getting a clearer starting point before I start exploring.
Tools: Miro AI (Sidekicks), ChatGPT, Figma Agents
Brief or no brief
AI pulls out the main points.
No brief? I jot my ideas on a board and ask AI to make sense of them.
Competitor analysis
Examples on a board. AI summarises them.
Summarising meetings
Notes on a board. AI summarises them. Copilot and Zoom AI can now do it straight away.
Competitor error pages on a Miro board, organised by error code, then summarised by Miro AI. One recommendation: a button that takes users back to the section they were in. I still make the design calls myself.
02 / 04
Ideation & exploration
This phase varies. I could be using Figma Make, Figma Agents, Claude, or none of them.
It tends to follow these steps
1
Explore: ask AI for 3 directions (usually Figma Make)
2
Review: flag mistakes, prompt back and forth
3
Add mine: my own ideas, plus what AI missed
4
Refine: feedback from the team at Design Surgery
5
Test: prototype it for user testing
Only step 1 is really the AI. Holding my ideas back at that point is on purpose: I want to see if it lands on the same thinking as me.
Dark mode toggle for Rayo web in Figma Make. I asked for three ideas across logged in, logged out, desktop, tablet and mobile, then kept prompting and added my own.
Newest: prototyping on our real components
We now have our own AI prototyping tool, built on our real components. A colleague in our team set it up, and I use it to prototype and test new ideas.
1
Prompt Claude for a new idea or version
2
It pulls in our real components from GitHub and our designs from Figma, via MCP
3
It builds a clickable prototype that looks and works like the live product
4
Share it with developers, or export it back into Figma, ready for user testing
My Library on Rayo web. Claude asked me two questions before building, then made a new version from our real components that I could click through and show developers.
03 / 04
User testing & production
AI helps with the summaries, but I still watch every usability testing video and build new components myself.
Where AI helps, and what I don’t skip
Usability testing
AI
Helps write the user test brief
AI
UserTesting’s AI summarises each test
Me
I still watch every video in full, so I don’t miss things and capture the user’s tone of voice
AI
Miro AI pulls out the quick wins
Production
Me
I build new components myself
AI
Figma Agents for the tricky ones, like breadcrumbs with three truncation rules
AI
Device images made in Adobe Stock Studio AI
Ways to listen page testing. Findings on a Miro board, then Miro AI pulls out the quick wins we can produce.
Those quick wins went to a Figma Agent, including copy updates, and I asked it to double-check structure and layout before production.
04 / 04
Presentations
Documenting recent work as decks, so we don’t forget it.
How I document my work
Me
Write the structure
AI
ChatGPT checks it
AI
A Figma Agent makes the slides from my template
Me
Edit it down and present it to the team
AI helps me work out what actually needs to be in the deck. Every call on it is still mine.
AI tools are great, but I use them as a supporting tool only.
It’s still on me to:
Understand the customer problem
Challenge what AI gives me
Back decisions up with research
Present and positively influence the people in the room
That’s what gets products over the line.
Sharing it