Trust
AI transparency
AI prepares, recommends, coordinates and executes bounded low-risk actions. Accountable people retain consequential decisions.
Where AI appears
Inside records at the point of work: summarising, drafting, qualifying and coordinating. Not as a detached chatbot making decisions.
Human gates
Every consequential action — client communications, commercial terms, publication, production changes — requires explicit human approval and is recorded.
Run register
Each AI run logs the actor, model version, inputs, sources, outputs and resulting actions, so behaviour can be audited and evaluated.
Red lines
No autonomous consequential decisions, no regulated advice, no personal or sensitive information in public consumer AI tools, no self-modifying agents.
Last updated 7 September 2026.
