Six rules you will be tempted to break.
They decide what to build, what to leave alone and when to stop.
The six principles
These decide what to build, what to leave alone and when to stop. They are not aspirations. They are the rules you will be tempted to break at the exact moment a vendor is showing you something impressive.
Workflow before automation. If the workflow is broken, automation only makes it fail faster. Fix the process first. Always.
Decide the outcome first. Agree what "better" means in hours saved, invoices collected or admin removed before anything is built. If you cannot name the number, do not start.
End to end, not task by task. Look at the whole journey from enquiry to delivery to follow-up. A fix in one step that creates a pile-up in the next is not a fix.
A human at every point that matters. AI supports people. It does not replace judgement. Every automation gets a named person who checks it before it reaches a client or makes a decision.
If the team will not use it, it does not count. A system that impresses on paper and sits unused has cost you money twice. Build what the people doing the work will actually run.
Build once, then keep it current. Design it so it still works when the business doubles. Then review it every quarter, because the tools change faster than the paperwork.
How to use them
Read them before phase one so the audit looks for the right things. Read them again before phase four, because that is where the pressure to skip them is strongest. If a proposed tool or automation breaks two of them, it does not go in, whatever it promises.
The one thing to do next
Go back to the AI Framework and start at phase one, or run the twelve-question assessment on the AI you already use.