Welcome to Module 4: Responsible Use of AI

Artificial Intelligence has become an integral part of modern workplaces, making tasks faster and easier. But knowing what AI can do is not enough. We must also know how AI should be used responsibly.

This module will guide you through the principles, practices, and mindsets needed to use AI tools safely, accurately, and ethically in your professional life.

![A person working on a laptop with a glowing AI icon, representing human-AI collaboration in the workplace]

Image Suggestion: A professional working at a desk with a laptop displaying an AI interface, symbolizing the partnership between humans and artificial intelligence in the workplace.

What We Will Cover

SECTION 1: FOUNDATIONS OF RESPONSIBLE AI USE

Why Responsible AI Use Matters

AI tools like ChatGPT and Gemini can produce output that looks highly professional. AI-generated text often has:

• ✅ Correct grammar

• ✅ Well-structured sentences

• ✅ Confident language

• ✅ Professional formatting

• ✅ Clear explanations

However, AI output can still be wrong!

AI may:

• ❌ Produce incorrect facts

• ❌ Misunderstand the context

• ❌ Invent information

• ❌ Use outdated information

• ❌ Present assumptions as facts

The Core Principle

AI should assist humans, not replace human responsibility.

AI can help prepare a first draft, analyse information, or identify possibilities — but the human user always remains responsible for deciding whether the result is accurate, appropriate, safe and suitable.

![A diagram showing the Human-in-the-Loop process flow]

Image Suggestion: A circular diagram illustrating the Human-in-the-Loop process: Human gives instruction → AI generates output → Human reviews → Human verifies → Human decides → Final action.

The Human-in-the-Loop Principle

A human remains involved in every important stage of the AI process — the AI never makes the final professional decision.

AI Can Assist — Humans Must Decide

A Real Workplace Risk

Professional Appearance ≠ Accuracy

Example:
An employee uses AI to create an official notice, and the AI gives an incorrect deadline. The message may look completely professional. But if it is sent to hundreds of employees or students, the mistake can create confusion and additional work.

Professional appearance does not guarantee accuracy. Responsible use requires human checking.

What Does "Responsible AI" Mean?

Responsible AI use encompasses seven key principles:

Responsible Use Protects:

1. Confidential Information — Prevents unnecessary exposure of internal data

2. Personal Privacy — Protects students, staff and clients at every stage

3. Accuracy — Reduces the risk of incorrect information through human checking

4. Human Control — Keeps humans in control of every important decision

SECTION 2: SIX COMMON AI MISTAKES TO AVOID

Overview

These mistakes may look small, but they can create serious workplace consequences.

1. Blindly trusting every AI answer

2. Accepting hallucinated information

3. Using vague prompts

4. Sharing private or confidential data

5. Copying AI output without proofreading

6. Sending AI-generated communication without a final human check

![An icon of a person with a question mark and a warning sign, representing common AI mistakes]

Image Suggestion: A visual showing a checklist with red X marks next to common mistakes like "Blind Trust," "Hallucinations," "Vague Prompts," and "Privacy Leaks."

Mistake 1: Blindly Trusting AI

"AI said it, so it must be correct." — This is not a safe approach.

AI generates responses based on patterns and available information. It does not automatically know whether a statement is true.

AI may confidently give:

• ❌ An incorrect date

• ❌ A wrong name

• ❌ An incorrect policy reference

• ❌ A wrong calculation

• ❌ An inaccurate statistic

Correct mindset: AI output is a draft or suggestion until it has been verified.

Mistake 2: Hallucinations

What is an AI Hallucination?

An AI hallucination happens when AI produces information that is incorrect, unsupported, or invented — while presenting it in a convincing way.

Examples of hallucinations:

• A non-existent policy

• An incorrect reference

• A fake citation

• A wrong date

• A non-existent person or source

• An incorrect numerical result

How to Reduce Hallucination Risk

We cannot assume AI will never make mistakes — but we can reduce the risk.

Safe Approach:
Ask AI to work only from the information provided or from an approved source, then verify the information against the actual official document — for example, the official university policy on late examination registration.

Mistake 3: Using Vague Prompts

A vague prompt increases the possibility of an unsuitable answer.

Vague Prompt:

"Write an announcement."

This does not tell AI:

• Who it's for

• What it's about

• What action is required

• The tone

• The length

• What must be included

A Better Prompt Provides:

• Role — Who should AI act as?

• Task — What should AI do?

• Context — What background should AI know?

• Audience — Who receives this?

• Tone — How should it sound?

• Requirements — What must be included?

Example of a Strong Prompt:

"Act as a university administrative assistant. Draft a formal announcement for faculty informing them about revised office timings effective from 5 August. Mention the new timings and ask staff to follow the revised schedule. Keep the notice concise."

Mistake 4: Sharing Private or Confidential Data

Employees work with sensitive information every day. The convenience of AI should never come before privacy and confidentiality.

Never enter into AI tools:

• ❌ Student records and employee information

• ❌ Personal identification details and financial information

• ❌ Internal reports and confidential institutional documents

• ❌ Passwords and login credentials

![A padlock icon with a shield, representing data protection and confidentiality]

Image Suggestion: A visual showing a person typing on a computer with a large red "X" over sensitive data like IDs, passwords, and student records, emphasizing what NOT to share.

Mistake 5: Copy-Pasting AI Output Without Proofreading

AI-generated text should never automatically become the final version.

Always Follow a Proofreading Process:

1. Generate → 2. Read → 3. Check → 4. Edit → 5. Approve

Mistake 6: Sending AI Emails Without Human Review

Final Checklist Before Sending:

• □

  Accuracy — Are all facts correct?

• □

  Relevance — Does it address the actual issue?

• □

  Tone — Does it sound appropriate?

• □

  Completeness — Is anything missing?

• □

  Privacy — Any unnecessary personal information?

• □

  Action — Is what's required clear?

• □

  Attachments — Actually included?

• □

  Final check — Ready to send?

SECTION 3: PRIVACY AND CONFIDENTIALITY

Personal vs Confidential Information

Data Minimisation and Anonymisation

Data Minimisation: Use only the information necessary for the task.

Replacements to Use:

• Student A · Employee B · Department X · Project Y · Client Z

Protecting Credentials and Sensitive Records

Never Share with AI:

• Passwords and OTPs

• API keys and authentication tokens

• Recovery codes

• Security answers

• Private access credentials

Before Uploading Student Records, Ask:

• □

  Is this information necessary?

• □

  Am I authorised to process it?

• □

  Is this AI tool approved for it?

• □

  Can the information be anonymised?

• □

  Can the task work without identifying details?

![A checklist icon with a magnifying glass, representing verification and review processes]

Image Suggestion: A visual showing a document with a magnifying glass examining it, with checkmarks next to items like "Names," "Dates," "Policies," and "Attachments."

SECTION 4: THE 10 GOLDEN RULES OF AI USAGE

Rules 1-5

Rule 1 — Always Verify Facts Before Sending

Check dates, names, numbers, policies and references.

Rule 2 — Never Upload Confidential Data

Protect institutional, student and employee data.

Rule 3 — Give Clear Role, Task and Context

The RTC framework improves usefulness.

Rule 4 — Ask for the Format You Need

Bullets, tables, checklists or step-by-step instructions.

Rule 5 — Break Big Tasks into Steps

Makes output easier to review and refine.

Rules 6-10

Rule 6 — Proofread Every AI Draft

Well written is not the same as factually correct.

Rule 7 — Keep a Reusable Prompt Library

Save prompts that work well for common tasks.

Rule 8 — Use AI as an Assistant, Not a Boss

Human = decision-maker, AI = assistant.

Rule 9 — Protect Student and Staff Privacy

Minimise and anonymise personal information.

Rule 10 — When Unsure, Ask a Colleague or IT

Never guess on privacy, security or policy.

What You Should NOT Do vs What You SHOULD Do

![An icon showing a person with a lightbulb representing human judgment and decision-making]

Image Suggestion: A visual showing a scale balancing "AI Suggestions" on one side and "Human Decision" on the other, with the human side weighing heavier.

AI as an Assistant, Not a Decision-Maker

Suppose AI says: "This employee should be denied leave." Should AI make that decision?

No.

AI can support decision-making, but sensitive decisions should not be blindly delegated to AI.

Human Judgment and Professional Responsibility

Human judgment considers:

1. Context — The broader situation and circumstances

2. Relationships — Professional and personal connections

3. Policies — Organisational rules and regulations

4. Ethics — Moral and ethical considerations

5. Consequences — Impact of decisions

6. Exceptions — Special cases and unique situations

7. Emotions — Human feelings and sensitivities

8. Responsibility — Ownership and accountability

SECTION 5: VERIFICATION AND HUMAN JUDGMENT

Verification: What Should Be Checked?

Always verify these elements before using AI-generated content:

Verifying Calculations, Dates and Deadlines

Calculations

• Ask AI to show the method

• Recalculate independently

• Use a calculator or spreadsheet

• Compare results before use

• Never copy a total straight into a report

Dates and Deadlines

• AI may misread "tomorrow" or "next Friday"

• Instead of vague language, give exact dates

• "The meeting is on 5 August 2026 at 3 PM"

• Specific dates remove ambiguity

• Always use exact dates for important work

![A calendar icon with a checkmark and a calculator icon, representing verification of dates and calculations]

Image Suggestion: A visual showing a calendar with a red circle around a date and a calculator displaying a sum, with a checkmark confirming accuracy.

Verifying Policies and Rules

AI is not the authoritative source of institutional policy unless verified.

Example:
"Employees are entitled to 20 days of annual leave" — must be checked against official HR policy first.

Correct Process:
✅ AI → Draft / Explain → Official Policy → Verify → Use

Unsafe Process:
✗ AI → Accept → Send

This distinction is extremely important in workplace communication.

Instructions That Reduce Risk

Use these instructions in your prompts to minimise errors:

• "Use only the information provided."

• "Do not invent missing facts."

• "If information is unavailable, say so."

• "Use placeholders where information is missing."

• "Do not change the meaning."

• "Do not add unsupported claims."

• "Separate facts from assumptions."

Safe vs Unsafe: Prompt and Email Examples

Result of Safer Approach:

• Explicit facts provided

• No invented details

• Privacy protected

• Clear instructions given

Responsible Use in Different Departments

Responsible Use in Research and Education

Research

AI Can Help:

• Brainstorm topics and summarise material

• Improve writing and organise ideas

But Verify:

• Citations, references, statistics

• Authors, publication details, quotations

• Never substitute AI for checking original sources

Education

AI Can Help:

• Lesson planning and activity ideas

• Explanations, summaries, feedback drafts

But Remember:

• Student privacy must be protected

• Teachers must check factual accuracy

• AI should not replace educational judgment

Responsible Use and Bias

AI systems can reflect biases present in the information used to develop or operate them — for example in:

• Language

• Recommendations

• Assumptions

• Descriptions of people

• Hiring-related content

• Evaluation criteria

When AI generates recommendations involving people, ask:

• Is the recommendation fair?

• Is there an unsupported assumption?

• Does the wording favour or disadvantage a group?

• Is the decision based on appropriate criteria?

![A balance scale icon representing fairness and bias awareness]

Image Suggestion: A visual showing a scale balancing different groups of people equally, symbolising fairness and the need to check for bias in AI outputs.

AI and Confidential Documents

Questions to Ask Before Uploading:

• □

  Is this document confidential?

• □

  Is this AI platform approved for this information?

• □

  Does the document contain personal data?

• □

  Can sensitive information be removed?

• □

  Is the task possible without uploading the whole document?

The Responsible AI Decision Test

Before using AI for any task, ask yourself:

SECTION 6: BUILDING A RESPONSIBLE CULTURE

A Simple Responsible AI Workflow

Follow these nine steps for every AI-assisted task:

![A flowchart showing the 9-step Responsible AI Workflow process]

Image Suggestion: A circular or linear flowchart with icons for each step: Define (magnifying glass), Minimise (scissors), Prompt (chat bubble), Generate (gear), Review (eye), Verify (checkmark), Refine (pencil), Approve (stamp), Use (send arrow).

Identify the Risk — Workplace Scenarios

Creating an AI-Safe Workplace Culture

Organisations Should Encourage:

• Clear AI guidelines and approved tools

• Staff training and privacy awareness

• Human review of outputs

• Clear escalation procedures

• Shared best practices

Employees Should Know:

• Which tools are approved

• What data may be entered

• What information is restricted

• When human approval is required

• Who to contact when uncertain

Responsible AI Maturity

Organisations and individuals progress through five stages of responsible AI use:

![A maturity ladder showing the five stages of Responsible AI Maturity from Beginner to Organisational]

Image Suggestion: A staircase or ladder graphic with five steps, each labeled with the maturity stage, showing progression from "Beginner" to "Organisational."

SUMMARY AND FINAL TAKEAWAY

Key Points to Remember

Final Takeaway

Use AI confidently, but never use AI blindly.

AI can generate the answer, but humans remain responsible for deciding whether the answer should be trusted, changed, shared, or used.

Quick Reference Card

The 10 Golden Rules (Quick Reference)

1. ✅ Always verify facts before sending

2. ✅ Never upload confidential data

3. ✅ Give clear Role, Task and Context

4. ✅ Ask for the format you need

5. ✅ Break big tasks into steps

6. ✅ Proofread every AI draft

7. ✅ Keep a reusable prompt library

8. ✅ Use AI as an assistant, not a boss

9. ✅ Protect student and staff privacy

10. ✅ When unsure, ask a colleague or IT

The Core Principle

AI should assist humans, not replace human responsibility.

Thank You

Knowspire × Almban

APPENDIX: ACTIVITY WORKSHEET

Identify the Risk — Practice Exercise

For each scenario, identify:

1. What is the risk?

2. What should the person have done differently?

Scenario 1

A manager asks AI: "Write a performance improvement plan for an employee who is consistently late." The AI generates a plan, and the manager emails it directly to the employee without review.

Your Analysis:

Scenario 2

An HR assistant pastes an Excel sheet containing employee names, salaries, and bank account details into ChatGPT, asking it to summarise the compensation structure.

Your Analysis:

Scenario 3

A professor asks AI: "Write a reference letter for my student who is applying for a PhD." The AI generates a letter with fabricated achievements and accolades.

Your Analysis:

Scenario 4

An employee uses AI to draft a response to a client complaint. The AI generates a polite response, but the employee sends it without checking the proposed solution against company policy.

Your Analysis:

Scenario 5

A team leader asks AI: "Which team member should I promote?" The AI analyses performance data and recommends a specific person.

Your Analysis:

Self-Assessment Checklist

After completing this module, you should be able to:

• □

  Explain why responsible AI use matters

• □

  Identify common AI mistakes and how to avoid them

• □

  Protect privacy and confidentiality when using AI

• □

  Apply the 10 Golden Rules in daily work

• □

  Verify facts, dates, and policies in AI output

• □

  Use the Responsible AI Workflow

• □

  Recognise when AI should not be used

• □

  Take responsibility for AI-assisted outcomes