AI Explorations
AI-driven product design, UX systems & creative automation
Overview
AI is no longer just a tool — it's part of how I think, design, and create. This section documents how I blend design thinking with artificial intelligence: from rapid prototyping to creative automation and visual exploration. These are not polished products — they're experiments, systems, and sketches that reflect a new kind of design process.
Lovable + ChatGPT — Prototyping at Speed
Working with Lovable and ChatGPT has completely reshaped how I approach prototyping and product ideation. With natural language prompts, I can generate fully functional app prototypes — not just mockups, but real interactive systems built on clean, production-ready React components.
I've created platforms ranging from nonlinear learning tools to dynamic dashboards, all without writing a single line of code. ChatGPT helps refine UX logic, clarify flows, and structure content, while Lovable instantly brings it to life.
These are not disposable prototypes — they're handoff-ready foundations. Developers can jump straight into Lovable's clean component structure, customise behaviour, and scale features without rebuilding from scratch. It bridges the gap between design and development.
Agent Spending Rules
AI Product Design · 2026
Claude Code — Agentic Spending Rules
Agent Spending Rules came from a shift that's already underway — AI agents are starting to buy things for us, with no click and no checkout screen.
The "are you sure?" moment doesn't disappear; it moves earlier, to a screen where you decide once what an agent can spend without asking you. That screen barely exists yet. So I designed and built it.
Working in Claude Code to define the decision logic and build the prototype in a single pass, I made an interface that lets you set your spending rules once or start from a profile, simulate a month to see where each purchase would land, and let anything unusual stop and ask — one at a time, and it tells you why. Everything the agent does is logged. You can pause or revoke in a tap. I then ran the concept through a simulated research panel and folded the findings straight back in: rule profiles to keep setup short, a cash-flow floor that guards timing and not just amount, a plain-language reason behind every decision. Concept to a tested, working prototype in a week.
This is what AI-native design makes possible — compressing the distance between an idea and something you can actually operate, test and react to, with design, research and iteration in the same place. Not a mockup. Not a concept. A real authorization layer, ready to sit inside a bank app or an agent and be taken further.
Lovable + Claude — Strategy Made Visible
FutureState came from a simple frustration — organisations struggle to commit to transformation because they can't see it yet. Strategy documents and slide decks don't make the future tangible. So I built a tool that does.
Working with Claude to define the product logic and Lovable to bring it to life, I created an experience-led transformation tool that lets teams visualise where they're going before a single resource is committed. The result is a working prototype built in days, not months.
This is what AI-native design makes possible — compressing the distance between strategic intent and something you can actually experience, share and react to. Not a mockup. Not a concept. A real, interactive foundation ready to be taken further.
Claude + Claude Design — TRADED, A Marketplace Rebuilt
TRADED came from a simple observation — classifieds platforms have quietly normalised injecting ads into listing grids, overwhelming users with filters, and blurring the line between sponsored and organic content. Users tolerate it because there's no alternative. I decided to design one.
Working with Claude to define the product logic and Claude Design to bring it to life, I built a full C2C marketplace from the ground up — starting with a structured audit of the category's failure points and designing every structural decision as a deliberate response to what incumbents got wrong. A homepage that serves two user modes simultaneously. A results page where sponsored content is contained, never injected. Filters that disclose progressively. A hierarchy that's honest throughout.
This is what AI-native design makes possible — compressing the distance between a competitive audit and a fully realised, interactive product. Not a wireframe. Not a concept. A working platform with a brand, a design system, and a complete user journey — ready to be taken further.
Iceland Winter Journey — Personal Travel Companion
AI Product Design · November 2025
Lovable + Claude — Iceland Winter Journey
A fully operational travel companion built for a 7-day winter road trip through Iceland — used every day, in the field, in the cold.
Built this for my wife and I ahead of a November road trip through Iceland — and used it constantly throughout the trip. Not a concept. Not a portfolio piece. A tool we actually relied on.
The app consolidated everything we needed into a single view: a day-by-day itinerary with expandable timeline cards, live aurora forecast with visibility windows, real-time road and weather conditions pulled from Icelandic sources, a trip currency calculator tracking daily spend, and offline mode for areas without signal.
The design challenge was environmental — the interface needed to work quickly, in low light, in gloves, often while one of us was driving and the other navigating. That shaped every interaction: large tap targets, high-contrast dark UI, collapsed-by-default detail, and a single-scroll structure so nothing was ever more than a swipe away.
What started as a practical planning tool became a live travel operating system. By Day 3 it had replaced every other app we'd opened.
"The best test of a product is whether you'd use it yourself. We used this every hour of every day."
Lovable + Claude — Meaning in the Age of AI
A personal project exploring where humans fit in an AI-driven world. An experimental platform designed to help people reflect on their strengths, values, and direction through interactive tests and reflections.
Rather than focusing on fear or hype, it explores how humans and intelligent systems can coexist and collaborate. Combines product design, AI experimentation, and philosophical inquiry.
Midjourney + ChatGPT — Visual Direction & Image Generation
Translating abstract ideas into meaningful visuals requires more than good prompting — it requires context, emotional intent, and the ability to iterate quickly. ChatGPT shapes the conceptual direction while Midjourney executes it with precision.
The result: a system for controlled randomness — shifting between raw expression and structured composition, generating images that reflect specific styles, emotional textures, and cultural cues.
n8n + Google Gemini — Creative Automation
Built fully automated creative workflows — from turning spreadsheet prompts into Midjourney-ready descriptions, to generating AI voices and triggering video content publishing. Modular, visual backend logic without code.
These workflows streamline design operations and make the invisible side of design visible, trackable, and repeatable.
"AI doesn't replace intuition — it sharpens it. The real power comes from how we choose to guide it, question it, and build with it."
Tools & Methods
Lovable, ChatGPT, Midjourney, n8n, Google Gemini, prompt engineering, rapid prototyping, creative automation.