Process

How I work, step by step.

I work methodically so that an intuition can be explained, tested and improved without losing its human side.

I do not follow these stages as a fixed recipe. I use them to free up attention: when repeatable work has a structure, I can spend more time listening, interpreting and making decisions that genuinely change the product.

Each stage leaves evidence and a rationale. That lets me discuss the work with others, return to it months later and be honest about what we know, what we are assuming and what still needs checking.

The six stages

Framing

Define the problem, the audience and the real constraint before drawing anything.

Human decisions: The problem, scope, success criteria and what is out of scope.
AI support: Gathering context and synthesising existing material.
Evidence: A written problem statement and acceptance criteria.

Research

Test my first impressions against the context and people’s needs.

Human decisions: What I need to learn, who to explore it with and which UX research technique could help.
AI support: Preparing discussion guides and organising authorised material, with source checks. This does not replace human participants.
Evidence: Questions, the chosen method and findings linked to their sources. I distinguish observations, interpretations and hypotheses still to be tested.

System

Define a foundation of typography, colour, spacing and components that can be maintained as the product grows.

Human decisions: Art direction, hierarchy, type scale and the character of motion.
AI support: Scaffolding variables and component states, and checking contrast.
Evidence: A collection of tokens and components with all their states defined.

Prototype

Screens assembled from system instances, navigable from day one.

Human decisions: Composition, flow and what to test before writing code.
AI support: Assembling variants, adapting to different screen sizes and documenting states.
Evidence: A navigable prototype that adapts to the agreed screen sizes.

Handoff

Maintain a clear relationship between Figma components and their implementation.

Human decisions: What belongs to the contract and what is left to implementation.
AI support: Reading the design structurally, generating components and testing.
Evidence: Code components that reflect the design system’s hierarchy.

Verification

Test the result in the browser and review what an automated test cannot assess.

Human decisions: The acceptable quality threshold and the debt to acknowledge.
AI support: Tests, accessibility auditing and performance measurement.
Evidence: A passing test suite, validated contrast and a design-to-browser comparison.

The role of AI

Making the boundary between judgement and execution explicit.

What the person decides

I define the problem, priorities and quality criteria. AI can suggest alternatives, but it is my responsibility to assess whether they make sense for the user and take ownership of the decision we make.

What AI carries out

I use it to explore variations, prepare components, review code and suggest tests. I give it context and clear limits, then compare the result with the design and the behaviour I expected.

How I keep it under control

I record what each tool can do, what needs my approval and how I will check the result. When a decision matters, I seek an independent review and retain the reason for the change.

Principles

System before screen

I want good decisions to be repeatable. That is why I connect screens to shared components, states and rules, and revisit the system when a new need arises.

Evidence over assertion

I try to show how I reached the result: the prototype, the components and the checks available. I also explain what is still in development or has not yet been validated.

Aesthetics serve a purpose

I pay attention to typography, spacing and motion because they affect how an interface is understood. If an effect distracts or a composition makes reading harder, I reconsider it even if I like how it looks.

Remove what does not help

I adapt the method to the problem. I document what helps us work and learn, without adding steps the team does not need.

See it in practice

In each case study, I explain what I was able to apply, the decisions I made and what still needs to be tested.

Explore the case studies