Designing in the age of AI.

AI in my design process

This is a pretty incredible time to be a designer. Everything is shifting fast, and rather than feeling overwhelmed by that, I find it genuinely exciting.

My workflow looks quite different than it did two years ago. I move between Figma and code more freely now, use AI at almost every stage, and keep adjusting as I learn what actually works. Here’s where it fits:

• Research and synthesis — I use Claude and ChatGPT to get through large amounts of information quickly: interview notes, competitor research, documentation. It frees me up to do the thinking that actually requires a person.
• Ideation — Good for getting unstuck and exploring more directions early on. I still make the calls, but AI gets me to more interesting starting points faster.
• Specs and PRDs — I cowork with Claude to turn rough thinking into product-requirements docs and thorough specs, structured and complete enough to actually build from.
• Placeholder copy — Copy that sounds real, so prototypes feel like the actual thing from the start.
• Prototyping — Figma Make in particular has changed how I prototype. I can test ideas much earlier, which changes the whole shape of a project.
• Design systems and build — This is the part I’m most excited about right now: using Claude Code to write directly into Figma files, build design systems, find problems and inconsistencies, and reorganise things at scale. It’s opened up work I simply couldn’t do before.
• Working in parallel — I’ve started using AI agents to work on multiple tasks at the same time. It’s a bit like having several versions of me running in parallel, each focused on a different problem. The amount of ground you can cover is hard to describe until you’ve tried it.

Made for AI

I got the chance to work on very exciting AI features that helped the sales team save time writing personalized cold emails at scale. Have a look →

Made with AI

I’m currently working on a set of AI-focused side projects to test new features, models, and to stay up-to-date with what’s possible. Adding them here as I go:

Friiiigo Live

A custom family calendar built as a single-page, no-backend web app, designed to be printed and stuck on the fridge. Users can add notes and highlight days before printing the desired date range.

I prototyped it in Claude using Sonnet, then moved into Figma to refine the design system: spacing tokens, a cohesive colour palette, and component variants for both landscape and portrait orientations. From there I brought it back into Claude to iterate on the colour scheme and sharpen micro-interactions throughout.

AI accelerated the parts that are typically tedious: generating the calendar grid logic, maintaining colour consistency across layout modes, rapidly testing palette variations, and enforcing print-safe styles.

The result, after a couple of days of work, is a lightweight, stateless app with no dependencies (just HTML, CSS, and a bit of JavaScript) that is going to massively improve our day-to-day.

Kian Flash Cards Coming soon

A native iOS flashcard app for learning languages, built with Expo and React Native. Features spaced repetition, a SQLite-backed card database, and smooth animated navigation.

I started by coworking with Claude on a product-requirements doc, then used Opus to sanity-check the planning and build the foundations. The fun was in the logic, spaced-repetition memory that keeps kids coming back, and a self-evaluation scale that stays gentle and hard to game (useful for kids who love to cheat, like mine). For the verb cards I tried Illustrator’s AI features first, but it couldn’t keep the vectors clean and consistent however clearly I briefed it, so I picked the work up where the AI left off, and still saved a good 50% on the manual drawing. For food, Claude tracked down a flat emoji set I could reuse. Thanks Claude!

This design still needs some love, but the app is fully functional, with 2 languages and 30 images per category. It’s incredible what I could achieve in a few hours.

Staying current

I follow researchers, practitioners, and critics across design, engineering, and ethics. I read documentation. I try things. I pay attention to what the broader conversation is missing, especially around accessibility, trust, and how real people actually experience AI-powered products.

Watching the pitfalls

When shipping gets this easy, the risk isn’t that you’ll move too slow. It’s that you’ll move too fast, on the wrong things.

AI can generate a lot quickly. But generating more has never been the hard part of design. The hard part is knowing what’s actually worth building, and staying honest about whether it serves real people or just feels satisfying to ship.

I’ve had to get more decisive about that. More opinionated, earlier. When the cost of execution drops, the quality of your judgment matters more than ever. Users don’t care how fast something was built. They care whether it works for them.