We use AI in our delivery. Here’s exactly how.

Our principles

Human responsibility, always

• A named engineer is responsible for every line of code that ships – whether it was typed or suggested.
• AI-assisted output goes through the same code review, testing and CI as hand-written code. Nothing bypasses that.
• Your data and your code stay yours. We do not train third-party models on your confidential information, and we configure our tooling accordingly.
• We’re transparent with clients about where AI is used in their delivery.

Where we use AI

Assistance, not autopilot

• Boilerplate and scaffolding – repetitive code that a person would otherwise copy-paste.
• Test generation, with a human choosing what to test and reviewing every assertion.
• Refactoring suggestions – evaluated against the codebase, not accepted blindly.
• Documentation drafts that our engineers edit for accuracy before publishing.
• Code review assistance – a second pair of eyes, not the only pair.

Where we won’t

Lines we don’t cross

• We don’t paste client code or data into general-purpose AI tools that could train on it.
• We don’t ship AI-generated code that nobody on the team has read and understood.
• We don’t use AI to make hiring, safety or clinically consequential decisions.
• We don’t dress up AI output as human work – clients know when AI has been part of the process.

Who owns the output

You do

The code we deliver is yours – the same as any other work we do. We keep records of how it was produced, and we can walk you through any part of it.