What people
decide.
An autonomous system needs a clear job and limits on what it can do. People define those limits and decide when to change them.
Start with the job.
We need to understand what the system is responsible for and how you will judge its work.
Some of that knowledge is already written down. Much of it comes from experience: knowing why one answer is acceptable and another is not.
We use those examples to define which decisions the system can make and which it should bring back to you.
Let it handle
the familiar cases.
Once a decision has been worked through, you should not have to make it again every time the same situation appears. The system applies what you have agreed and checks what happens.
The system operates.
The human steers.
The results matter. If an approach stops working, the system needs to recognize that. People decide whether the goal or the rules should change.
Keep reviewing
the rules.
A rule that worked last month may be wrong today. A system can hit a numerical target while missing the reason you chose it.
You need to be able to see why the system made a decision and correct it when necessary. We treat that review as part of operating the system.