Process

My design process is a classic double diamond with a discovery phase where team agrees on direction, and an iterative design phase where team signs off on a solution. User research runs parallel, end to end.

Discovery

Tech stack and product roadmap ground discovery. Research is done in the market and around our audience: competitive landscape, best practices, and products built for our ICP. It shapes candidates tailored to our target, which are reviewed with team to pick a direction for design iteration.

Design

After discovery, rounds of design iteration are run with team and stakeholders. Designs are refined with each iteration until they can be tested before delivery.

  1. Iteration starts from the direction agreed in discovery, with constraints already mapped.
  2. Design iterations run in rounds of team review and feedback, sharpening the work until everyone agrees to move on.
  3. The agreed direction is put in front of real users and tested before being shipped.
  4. Refined designs are then delivered to engineering, with design support through build and QA.
User research

Research runs alongside the process: at each stage there are methods that can resolve the team's doubts with real user feedback. Where there are none, established patterns can be enough.

  • Ideas & concepts / Interviews, concept tests, card sorts and tree tests — what problem are we actually solving, and is this worth building at all?
  • Content / Copy tests, five-second tests, comprehension and naming — does the wording land the way we mean it, and do our words match theirs?
  • Design & prototype / First-click and preference tests, design surveys, live prototype tasks — do they start in the right place, and can they complete the journey?
  • Before release / Live website tasks, usability passes, accessibility checks and unmoderated studies — does it hold up on the real build, and does it work for everyone?
The process with AI

Design work has been transformed with the advent of powerful generative AI tools. I have successfully automated my full process with AI. The main principle behind it is to replicate traditional product workflows and roles with a modern stack of tools and coding agents.

AI-powered discovery

Market and competitive intelligence, user journeys, ux patterns and modern best practices are scanned and synthesised with agentic workflows to validate a product direction. Insights shape decisions within hours rather than weeks.

AI-powered design

The AI assisted pipeline replicates a traditional iterative design process, partially or fully, but the user is the stakeholder and the AI coding tool is the designer.

  1. The stakeholder makes a request using the orchestrator skill /ux
  2. AI coding tool responds by orchestrating agents that ideate iteratively based on stakeholder feedback
  3. Real stories are created and reviewed by a fidelity agent and the stakeholder
  4. A signed off story comes back as components, mapped to the design system and ready to test and build.
AI user research

User research should not be a blocker in AI accelerated product workflows. I built fiuto.ai as a personal tool to solve for this mismatch. I've since built it into a real product.

See the full portfolio presentation The process, selected projects and how I work with AI (opens in a new tab)