How I use AI models and agents in client work
Claude, ChatGPT, Copilot, Cursor and Codex speed up delivery. Here is where they help, and where a person still decides.
I use AI models and coding agents every day: Claude, ChatGPT, GitHub Copilot, Cursor and Codex. They are part of how I deliver 50+ web and eCommerce projects without cutting corners on quality.
Where AI helps
- Drafting boilerplate, tests and migration scripts so I can spend time on the hard parts.
- Exploring an unfamiliar codebase or a legacy plugin before I touch it.
- Turning rough client notes into a structured brief and a first-pass scope.
- Reviewing my own work for edge cases I may have missed.
Where a person still decides
Architecture, security, data models and anything that touches a client's customers. Generated code is read, run and tested like any other code, and nothing ships that I cannot explain.
Building the bridges myself
I also build the tooling that connects products to AI agents: Model Context Protocol (MCP) servers and AI-integrated plugins with multi-provider support. That lets a WordPress site or a SaaS product take part in an agent workflow in a controlled, reviewable way.
What clients get
Faster delivery, more time spent on the decisions that matter, and one person accountable for the result from scoping to production.
- AI
- MCP
- Workflow
