An AI policy is the organisation's shared ground rules for using artificial intelligence: which tools are approved, what they may be used for, what must never be shared, who is responsible – and how to be open about the role AI plays in the work.
Most AI policies do not fail on content but on form: they are written as compliance documents and never read. A policy that works is short, concrete and written in everyday working language – and it takes shape in dialogue with the people who have to live by it.
The policy is also a strategy document in miniature: it reveals what the organisation believes AI is for. That is why it should begin with the purpose – what we want to achieve – before it moves on to the limitations.
In practice
An AI policy that fits on two pages gets read: approved tools, golden rules for data, a requirement for human approval where it matters, openness about use – and one place to ask when in doubt.
How to explain it to leadership
»An AI policy is not a restriction on the work – it is a shared agreement on how we want it done.«
The ground rules are typically set at my AI strategy workshop – and embedded through the adoption work.