Messaging at scale
A production Pub/Sub service that processed 9.2 million HTTP requests, including message responses and status updates, in a 30-day period.
Next case studyAn AI-first way of working
A team guide for working AI-first: plan first, test first, and review every change twice.
- Role
- Wrote the guide and set the approach
- Client
- A business-messaging platform
- Scope
- How the team plans, builds and reviews with AI coding tools
- Review
- Automated and human, on every change
01
The problem
AI coding tools let a team produce far more code, and far more code to review. Without clear habits, speed turns into rework and low-quality output.
02
What I did
I wrote the team’s best-practice guide for working AI-first and set the approach: a written plan agreed before any code, tests written first, and an automated AI review plus a human review on every change before it merges. The guide also covers what to hand to the tools and what to keep in human hands.
Diagram of the way of working, drawn as a loop. A change starts with a plan, then an AI draft, then tests. It goes down to an AI review, then a human review, and then it merges. A dashed line leads from the merge back to the plan for the next change.
03
Results
Planning
Plan first
A written plan agreed before any code is generated.
Testing
Tests first
Tests describe the behaviour before the code is written.
Review
Two reviews
An automated AI review and a human review on every change before it merges.
04
Stack
- Claude Code
- automated code review
- Pytest
- Playwright
Contact
Get in touch
Tell me about your project, your team or the problem you're working on.
Or find me on LinkedIn