Method in detailMethod in detail
The method, from discovery to measurement
A human-centred method, for solutions that are desirable (design), viable (business) and feasible (tech). Discovery and delivery move forward in parallel, without slowing the team. For each phase, the tools, including those that use AI. AI does two things: reach a testable prototype faster, and design useful AI features. Decisions stay with the team and the users.
01
Continuous discovery
Find real, high-impact problems, not features to ship.
AI Gemini and custom Gems to surface patterns, Condens to centralise and analyse interviews.
At Decathlon Gemini and custom Gems to cross verbatims with usage data, Perplexity for desk research. Trends emerge faster and the team decides on solid ground.
02
Definition
Frame the right problem, then prioritise it with the PM and lead tech.
AI Claude to structure the Opportunity Solution Tree from insights, and prepare impact / effort prioritisation.
03
Framing
Align the trio on stakes, assumptions, constraints and uncertainty, then pick the right method.
AI A quick prototype made with Figma Make to defend the project at the IPR committee and unlock the budget, as at Leroy Merlin. Perplexity to enrich briefs and challenge assumptions.
04
Further research
Feed the thinking with what exists elsewhere.
AI NotebookLM as an external brain: sourced syntheses of interviews and studies.
05
Ideation
Explore widely, then converge together. The designer facilitates rather than designing alone.
AI Claude to prepare workshops (“How might we” questions, timed agenda), Figma Make to explore several directions in minutes.
At Leroy Merlin Designing for AI: Copilot RO's first AI feature, in preparation, and DiagAssist, an AI assistant for customer service. Useful, understandable AI that users stay in control of.
06
Prototyping and testing
Test fast, learn fast, with the design system and accessibility from the start.
AI Figma Make, GitHub Copilot and Apps Script: prototypes coded on the design system, with realistic data, tested before development. Stark for accessibility.
At Leroy Merlin Prototypes coded with Figma Make, GitHub Copilot and Apps Script, directly on Mozaic, the group's design system, fed with realistic test data. We test with users before the first line of development.
07
Delivery
Design built into the sprint, all the way to a faithful release.
AI Claude Code and the Figma MCP for complete handoff specs and components coded on the design system.
08
Measurement
Prove value through usage, and find the next opportunity.
AI Gemini to cross usage data with customer feedback, Claude to prepare the impact presentation for decision-makers.