How we use AI
The rules we hold ourselves to.
A practice that advises on AI should say how it uses AI. These are the rules in our own policy, in plain language.
In client work
Six rules.
Approved tools only
We keep a list of AI tools approved for client work. Anything not on it is not used. General web chat tools are not approved.
Client data is processed locally
AI tools write the scripts. The scripts run on our own equipment against your data. The contents of your files are not pasted into prompts.
No retention by the provider
The AI accounts we use for client work run under terms that stop the provider keeping what passes through.
No client data in code repositories
Only code is kept under version control. Data files never are, in public or private repositories.
AI output is untrusted until checked
Code and documents produced with AI go through the same review as anything written by hand, at a level set by the cost of a mistake.
A person is accountable
Every deliverable has a named person who has read it and stands behind it.
On this site
How this site was made.
This site was designed and built with the help of AI tools, and reviewed by a person before publication. The copy draws on articles, talks and project records written by our own people.
The readiness check does not use an AI model. It is scored by ordinary code, so the same answers always give the same result.
There is no chat assistant on this site. When you contact us, a person replies.
If we fall short
Tell us, and a person will answer.
These rules are reviewed as the tools change. When we find a gap, we change the rule and tell the clients it affects.
If you believe we have fallen short of them, email hello@sas-am.com. It will be read by the Managing Director.