Responsible AI
We use AI as a tool inside our own delivery, not as a product we sell you. This is how we handle it — the boundaries, who stays accountable, and where it sits in our existing controls.
We use AI inside our own delivery — retrieval over your documentation, first-draft user stories from your specs, faster onboarding to an unfamiliar codebase. It speeds up how a senior engineer works. It does not design your systems, and we do not build AI features into your product unless that is the engagement you have asked for.
Where the work touches your code or internal documentation, we use local, self-hosted models — the material is not sent to a third-party AI service, and it is never used to train a model. When a hosted model is genuinely the right tool for a task, we agree it with you first and scope exactly what it may see.
Model output is a first draft, never a decision. A senior engineer reviews it, corrects it, and owns what reaches your codebase — and it passes the same test-effectiveness measurement and adversarial review as any other change. The model is held to your quality bar, not exempt from it.
This sits inside the same management systems as the rest of our delivery — ISO/IEC 27001:2022 for information security and ISO 9001:2015 for quality — and every engagement is covered by a confidentiality agreement before any sensitive scoping.
If your security or procurement team has questions about how a pod would handle AI on your codebase, we will answer them in writing before any engagement begins.
Start a conversation →