Develop with AI. Do not delegate the project to AI.
AI acts as an augmented working environment. It helps explore several options, understand an existing codebase, produce a first version and document decisions more quickly.
I define the need, architecture and priorities. I control the data, review the code, test the journeys and validate every delivery. The project remains guided by real use, not by what the tool happens to make possible.
What AI genuinely contributes.
Used methodically, it reduces time spent on mechanical tasks and creates more room for product quality.
Explore faster
Compare architectures, identify constraints and assess several solutions before choosing.
Prototype before investing
Turn an idea into a testable journey and validate its use before full development.
Automate without rigidity
Remove repetitive operations while keeping business rules understandable and editable.
Document as the project grows
Produce technical and functional documentation that supports adoption and future changes.
Audit in greater depth
Review an existing system, find inconsistencies and prepare improvements without needlessly starting over.
Evolve sustainably
Accelerate fixes and new features without sacrificing clarity or maintainability.
Four stages. One clear responsibility.
Scope
Objectives, users, data, constraints and success criteria are defined before production starts.
Design
Journeys, architecture and prototypes are designed and checked against the real need.
Build and verify
Code is reviewed, tested, corrected and challenged against use cases, errors and limits.
Deliver and transfer
The solution is documented, explained and prepared to remain understandable and adaptable.
Technology assists. Expertise decides.
What remains human
- Business scope and priorities
- Architecture and data model
- Security, testing and validation
- Editorial direction and user experience
- Trade-offs and final accountability
What AI accelerates
- Analysis and option exploration
- Code suggestions and refactoring
- Test scenario generation
- Inconsistency detection
- Documentation and repetitive tasks
Codex and Claude, selected for the context.
These tools replace neither a method nor business expertise. They are selected for complementary strengths and used within a clearly defined framework.
Codex
A development agent for analysing codebases, implementing, refactoring, running tests and documenting work in a controlled environment.
Claude Code
A collaborative environment particularly useful for understanding a project, reasoning about its architecture and supporting complex changes.
Start with the problem to solve.
I help determine whether AI brings a genuine benefit, then build a clear, tested and maintainable solution.
Jonathan BricourtDigital · Photographie