5.5AI usage charter
The essentials in 30 seconds
- One page, five sections: tools, data, checking, transparency, responsibility.
- The middle way: neither total ban (shadow AI) nor laissez-faire (incidents).
- Living charter: approved, shared with discussion, reviewed on a fixed date and fed by grey cases.
Running case — Amine gets the one-page charter signed
Amine compiles his whole journey in Algiers: decision tree, 3 data rules (5.1), checking protocol (5.2), images rule and DPO radar (5.3-5.4). He gets a one-page first draft written by AI, confronts it with 3 real grey cases from the team, gets it approved (manager, IT/DPO) and sets the review. Last episode of cases 5.1-5.4 — Amine turns his reflexes into a team framework.
Last episode of cases 5.1-5.4 — Amine closes the level with the team deliverable.
1. Five sections
The guide's culmination: the usage charter, the one-page document turning individual practices into a team framework. Five sections: authorised tools (and forbidden ones), data (what may/may not be shared), checking (who approves what), transparency (when to flag AI) and responsibility (whoever signs takes ownership). Amine pours into it each learning from previous sections.
2. The middle way
A realistic charter: neither total ban (which pushes towards shadow AI) nor laissez-faire. Build on the whole journey: decision tree (4.5), confidentiality rules (5.1), checking protocol (5.2). It is a living charter: reviewed on a fixed date (tools change fast) and fed by grey cases met.
An imposed charter stays dead letter: confront the first draft (written by AI itself) with 3 real grey cases, get it approved at the right level, share it with discussion time — then keep the “AI case of the month” slot in meetings.
🔎 Concrete example — Bringing the charter to life
An imposed charter stays dead letter: Amine's method.
Get the 1st draft written by AI, confront with 3 real grey cases, adjust.
Approve (manager, IT/DPO), set the review date.
-> release with discussion time: a charter is lived, not imposed.
Amine adds the “AI case of the month” slot in team meetings.Outcome: grey cases make the charter's richness, not general principles.
Lesson: a monthly slot anchors the charter in practices.
Section vocabulary
- AI usage charter
- The one-page framework: tools, data, checking, transparency, responsibility.
Simple example: Amine keeps his signed team charter known to all.
What it is not: a PDF signed once then forgotten in a folder.
Reflex: does everyone on the team know the charter?
- Tool allowlist
- The approved tools per data level: the operational heart.
Simple example: Amine reads in it which tool for which data level.
What it is not: “use whichever tool you prefer” with no list.
Reflex: is the list up to date and accessible?
- Data rules
- Classification translated into simple team moves.
Simple example: Amine pours his 3 personal rules from 5.1 into it.
What it is not: general principles with no team examples.
Reflex: would a newcomer know what to paste where?
- Human approval
- Who re-reads what before release: explicit circuit.
Simple example: Amine writes his 5.2 checking protocol into it.
What it is not: “we re-read when we have time”.
Reflex: who signs what, and after which check?
- Living charter
- Reviewed on a fixed date and fed by grey cases met.
Simple example: Amine schedules the review and the monthly slot.
What it is not: a frozen charter while tools change.
Reflex: is our next review dated?
In practice — Writing your team's one-page charter
- Compile your learnings: decision tree (4.5), confidentiality rules (5.1), checking protocol (5.2), regulatory points (5.4).
- Get the first draft written by AI itself: “Write a one-page AI usage charter for a [context] team: authorised tools, data rules, checking, transparency, responsibility. Clear, applicable, non-legal tone.”
- Confront it with 3 real grey cases from your team: does the charter answer them? Adjust.
- Get it approved at the right level (manager, IT/DPO for sensitive points) and set the review date. Release with discussion time — a charter imposed without discussion stays dead letter.
Ready-to-say script — getting the charter approved
“Here is our one-page charter: approved tools, data rules, checking, transparency, responsibility. It answers 3 real grey cases from the team. Do you approve this framework? We will release it with discussion time and review it on the written date.”
To say to the manager then to IT/DPO for sensitive points, one-page charter in hand.
Copy-paste template — one-page charter
AI USAGE CHARTER — TEAM: _______________ (date: ___ / review: ___)
1) TOOLS: authorised: ___ | forbidden: ___ | Allowlist kept by: ___
2) DATA: shareable: ___ | never outside approved tool: ___ (see 5.1)
3) CHECKING: who approves what (5.2 protocol): ___
4) TRANSPARENCY: when to flag AI: ___
5) RESPONSIBILITY: whoever signs takes ownership: ___ | Questions: DPO/contact: ___Solo / freelancer
Amine solo: his charter fits in 5 personal lines + tool list — same standard, lighter format.
Small business / small team
One page posted, 3 grey cases tested, manager approval: zero heaviness, clear framework.
SME in Algeria
Amine in Algiers: bilingual charter if needed, real local cases, IT/DPO approval on sensitive points.
Growing team
“AI case of the month” slot in meetings + question channel: each newcomer signs after discussion, not before.
✓ Best practices
- Keep the charter to one page, five sections.
- Aim for the middle way: neither total ban nor laissez-faire.
- Set a review date (the AI landscape changes fast).
✗ Common pitfalls
- Total ban (pushes towards shadow AI).
- Laissez-faire (leads to incidents).
- A charter imposed without discussion (stays dead letter).
Key takeaways
- One page, five sections: tools, data, checking, transparency, responsibility.
- Total ban pushes towards shadow AI; laissez-faire towards incidents.
- Grey cases met make the charter's richness: collect them.
- Dated review mandatory: the AI landscape changes fast.
48h challenge — apply tomorrow
Compile your 5.1-5.4 learnings, get the one-page first draft generated and confront it with 3 real grey cases: approval planned?
- I can cite the five sections without hesitating.
- My one-page first draft exists (grey cases tested).
- Approval and dated review are planned.
Frequently asked questions
Our company already has an AI policy: why a team charter?
Group policy sets the framework; the team charter translates it into concrete moves for YOUR tools and YOUR use cases. One does not replace the other — the team charter cites and applies the policy.
How to keep the charter alive beyond signing?
Three levers: an “AI case of the month” slot in team meetings, a channel to ask questions, and the announced review. The charter lives if people talk about it.
What are the five sections of an AI usage charter?