AI can make suggestions and execute tasks. Domain and technical decisions remain with the people who are accountable for their consequences.
Our position
Manifesto for AI in Software Engineering
AI is not a substitute for engineering discipline. It amplifies the structures already present in a system—good and bad alike. We therefore combine its use with clear accountability, verifiable quality, and architecture that provides direction.
Shape AI adoption in your organizationGuiding principles
How we judge AI-assisted development
Value does not come from the tool alone, but from a way of working in which people understand, verify, and take responsibility for the results.
Clear modules, stable interfaces, and dependency rules constrain what AI can change and protect the integrity of the system.
Tests, reviews, static analysis, and CI rules apply equally to changes produced by people and machines.
Code must remain understandable, maintainable, and verifiable. Short-term throughput must not create long-term uncertainty.
Tools receive only the information required for a task. Privacy, confidentiality, and security are part of the design.
We use AI where it improves feedback, reduces risk, or demonstrably saves time—not because using it appears modern.
From principles to practice
Guardrails must work in daily development
Teams need more than a policy. They need to test on realistic tasks which work to delegate to AI, how to provide context, and how to safeguard the results.