That’s a thoughtful question — and one many people, organisations, and governments are actively working through right now. Since none of your personal documents, emails, meetings or people records contained material on this topic, here’s a clear and balanced overview to help shape your thinking.
What AI should do
These are areas where AI tends to be beneficial, low‑risk, or easily supervised by humans:
Assist with repetitive, time‑consuming tasks, such as summarising information, organising data, scheduling, drafting content, or checking for consistency.
Support with decision‑making, as long as a human stays in the loop — for example, helping compare options, surfacing patterns, or highlighting risks.
Improve accessibility, like real‑time captioning, language translation, or personalised learning support.
Enhance creativity and productivity, by helping brainstorm ideas, generate images/text, or automate parts of creative workflows.
Monitor for safety and quality, like spotting anomalies in systems, detecting fraud, or improving cybersecurity.
What AI should not do
These are areas where giving AI the authority or autonomy may introduce disproportionate risk or ethical concerns:
Make final decisions that affect people’s rights or wellbeing, such as hiring, firing, medical diagnoses, legal judgements, credit approvals, or policing actions.
Operate without oversight in safety‑critical contexts, such as autonomous weapons, infrastructure control, or medical treatment.
Handle highly sensitive personal data without explicit consent and robust safeguards.
Generate or spread disinformation, impersonate individuals, or manipulate emotions at scale.
Replace human accountability — AI is a tool, and responsibility must remain with people.
A helpful framing A good rule of thumb:
AI can analyse, assist, accelerate or advise — but humans should decide, direct, and remain accountable.
It looks like there are no AI‑related internal guidelines in your organisation’s files, emails, meetings, or people records based on my search (I checked across all domains) 1234567.
So here’s a practical, ready‑to‑use framework you can apply directly in your organisation to turn the principles we discussed into actionable internal practice.
How to Apply AI‑Use Guidelines in Your Organisation
1. Start with a simple, shared AI policy
You don’t need something complex at first. A concise, accessible policy should state:
What AI can be used for (e.g., drafting, summarising, brainstorming, translation)
What AI cannot be used for (e.g., sensitive decisions, personal data without approval)
A reminder that humans remain accountable for all outputs
I can help draft a first version if you like.
2. Define “human in the loop” checkpoints
AI should assist, not autonomously decide.
For every workflow where AI is used, specify:
What the AI does (e.g., summarises customer feedback)
What the human does (e.g., reviews for accuracy and fairness)
What risks to check for (misinformation, bias, hallucinations)
This protects both staff and operations.
3. Categorise tasks into “green”, “amber”, and “red” zones
This makes adoption easier and reduces uncertainty.
Green zone (safe to automate or assist):
Drafting emails, documents, training materials
Idea generation
Summaries of non‑sensitive content
Admin tasks (scheduling, formatting)
Amber zone (allowed with oversight):
Analysing trends from internal data
Customer communications
Recruitment content (e.g., job posts)
Requires human review + approval.
Red zone (not allowed):
Assessing candidates
Making performance or HR decisions
Medical/financial/legal determinations
Handling confidential or regulated personal data
Anything safety‑critical
4. Build AI literacy across the organisation
This increases responsible use naturally:
Short monthly learning sessions
Examples of good vs. risky AI usage
Quick guides or checklists
Show staff how to double‑check AI outputs
This also reduces reliance on informal, inconsistent practices.
5. Appoint an “AI steward” or small committee
They don’t need to be AI experts — only responsible for:
Reviewing ambiguous cases
Updating guidelines
Helping teams adopt AI safely
Monitoring compliance lightly, not punitively
This ensures oversight without bureaucracy.
6. Integrate AI guidance into workflows and tools
For example:
Put reminders in templates: “Did you check the AI’s output for accuracy?”
Add automatic warnings when handling sensitive data
Include quick links to approved AI tools
This makes the policy live in everyday work.
7. Review and update regularly
AI changes fast, so treat policies like living documents:
Quarterly reviews
Collect examples of misuse and best practice
Ask staff what’s working and what isn’t
Iteration is key.
If helpful, I can create:
A draft AI policy tailored to your organisation
A training slide deck
A Green/Amber/Red reference chart
A decision checklist people can use before involving AI
Would you like one of those?
7 H H 