9order LEARN 9order ↗
Learn / AI Operating Charter

AI Operating Charter

Record your question, keep your work and explain the evidence behind your next step.

Workbook / self notes Self notes · not a prompt
What you'll learn

Use this charter to record ownership, evidence, boundaries, failure handling, quality, value and release decisions for an AI capability.

Bring / required

No AI is required to fill in this page. Bring your own service/workflow notes, or use the starter case. Use AI/GPT only when an activity is explicitly marked Prompting exercise.

What you'll do

Record your opening question, update the operating charter during the course and distinguish observed evidence from assumptions.

Take away

Your AI Operating Charter, current operating decision, evidence gaps and next review trigger.

This is a working record, not a prompt and not proof that a production AI system is safe or ready.

How to use your workbook

This page is for your own notes and evidence. It is not a prompt library. Write in the cards, save on this device, copy, print or download. When a course activity wants you to paste something into AI/GPT, that activity will be explicitly labelled Prompting exercise and the copyable text will be separated from explanation.

My question and what I learned

Self notes · not a promptWrite your own question and what would count as a useful answer. Keep this as your operating record; do not paste the whole card into AI.
MY QUESTION
Course / date:
My work or curiosity:
My priority question:
My route: own work / help choosing / starter case
A useful result would be:
Experience / access help needed:

RETURN TO MY QUESTION
Status: answered / demonstrated / needs more evidence / outside this course
I came in asking:
My answer now:
The output, decision or evidence that changed my view:
What I still need to find out, and how:
My next small application:
Optional next-week check: new input / what held / what changed

AI Operating Charter

THE CAPABILITY
Useful outcome / users / limits:
Business owner / operator / exception decision maker:
Current mode and reason:
Allowed actions / approval / stop rule:
Test cases: input / expected behaviour / observed result / decision:
Critical failures that block release:
Failure signal / next check / fallback / recovery owner:
Record needed to reconstruct what happened:
Health sign / threshold / action / owner:
Quality sign / threshold / action / owner:
Value and cost unit / evidence / unknowns:
Change that requires re-testing / affected cases / release owner:
Current decision: proceed within stated scope / add controls / reduce authority / stop
Evidence type and remaining gaps:
Next review trigger or date and owner:
Question status and next application:

A completed example

Fictional worked example. Inspect it, then record your own reasoning and evidence.

  • Question: Would I allow this quotation assistant to send on its own?
  • Owner/operator: Mira owns business outcome; training operator investigates execution; Farah handles delivery follow-up.
  • Decision: Prepare mode only; checked human sending. T3's unauthorised send and S10's stale price block automatic sending.
  • Recovery: Hold uncertain retries, inspect sent records, assign missing follow-up, retain approval/version/receipt together.
  • Quality rule: Any unauthorised send or incorrect price triggers restricted sending and affected-case recheck by Mira. This is a classroom rule, not a universal acceptable-error standard.
  • Value: RM1,500 nominal time value versus RM1,600 stated operating cost gives −RM100 before other benefits/costs; realised cash savings and sales uplift are unproven.
  • Release: Retest new price, missing details, discount refusal and approval before sending after changes. Mira reviews before increasing authority and after any critical failure.
  • Evidence: Simulated operating decisions; no claim of production readiness.

What is it allowed to do?

Mode What may happen Quotation example
Observe Read allowed information and report it Identify a missing quantity
Recommend Propose a next step; person decides and acts Recommend asking for delivery details
Prepare Create a draft for review; no customer send Draft Q-104
Act after approval Perform a specified action after the named approval Send the approved version to the verified recipient
Act within limits Perform only a pre-authorised class of actions within tested limits Update a permitted training status field; exceptions go to the owner

Ask separately what a mistake could change and whether it can be reversed. Writing a rule is not proof that a system enforces it.

Check request and facts → confirm authority or obtain approval → act → verify the resulting state.