How we use AI
Where AI enters our work, what it can and cannot see, and how we keep people in control of every decision.
On this page
We use AI every day, in design, in code and in operations. It makes us faster, and it lets a small, senior team take on more than it otherwise could. Decisions and responsibility stay with people, and your data is protected with the same care we apply to everything else.
Where we use it
In prototyping and design, in writing and reviewing code, in documentation, and in analyzing infrastructure and logs. The tools we use today are Claude Code and Gemini. If that list changes, we tell you before the change reaches your project.
Privacy and data sharing
- Your project’s source code is processed by these tools during development.
- Production data and your users’ personal data do not go into AI tools without your written authorization.
- Passwords, keys and credentials never do.
- We use only commercial plans whose terms prohibit training models on your content.
How we keep AI within limits
We work with AI the way we work with any other engineer on the team: with clear instructions, a defined scope and review at every step.
- Specifications first. Before any code is written, the work is described in a written spec, and the AI works from that spec instead of guessing what we want.
- Short steps. Work is broken into small, reviewable steps. Each step is checked before the next one starts, so a mistake never compounds.
- Written project rules. Each project has documentation that tells the tools how the codebase is organized, which conventions to follow and what they must not touch.
- Minimal access. Connections between AI tools and other systems (through MCP) are set up with the minimum access the task requires, and nothing beyond it.
- A full record. Every change goes through version control with a description of what was done and why, so anything can be traced and reviewed later.
People decide
Code produced with AI goes through the same review as code written by hand before it reaches production. Decisions about architecture, security and data are made by people.
Ownership
What is produced with AI in your project follows the same rule as everything else: it is yours once the project is paid in full.
Transparency
At the start of every project, we tell you which tools we use and how. If you would rather we not use AI in some part of the work, we agree on that in the contract.