Real AI confidence, one small step at a time.
A guided playground for MTN people. Learn to use AI safely, get real value in everyday work, and build confidence through practice, not theory.
Level 0: Start Safe
Know what is safe to share and which tool fits which task.
Six levels, from safe first steps to leading your team
Each level is short, practical and ends with a five-minute mission. Level 0 is mandatory for everyone.
Earn a badge for every level you complete
Momentum you can see. Complete a level's missions to unlock its badge.
🔥 Momentum: 1 day. Small and often beats rare and long.
Toolkits you can open any time
Practical tools, not just links. Keep them handy while you learn and while you work.
Your Level 0 to Level 5 pathway
Follow it in order for the smoothest ride, or jump to what you need. Level 0 comes first for everyone, it keeps you and MTN safe.
AI will not simply take your job. People and teams who use AI well will outperform those who do not. This academy is how you get on the confident side of that line, safely and at your own pace.
Start Safe
Before you use any AI tool, learn what is safe to share, which tool fits which task, and where the human stays accountable. This builds on MTN's Responsible AI programme.
The safety foundation: what you can and cannot put into AI, and which tool to choose.
Used carelessly, AI can expose customer, financial or confidential data. A few simple rules keep you safe.
Read the four rules, run the Tool Safety Selector, then classify one real task.
The task classifier below. It checks your task before you share anything.
Finish MTN's Responsible AI programme before you begin the academy. It covers accountability, fairness and governance. This level is your practical companion to it.
Why Copilot is the MTN standard
MTN's organisation-wide standard is Microsoft Copilot. It has enterprise data protection, which means your prompts and company data stay inside MTN's protected environment. Start here for anything involving MTN information.
If the task touches MTN data of any kind, use Copilot. Public tools like ChatGPT, Perplexity, Claude, Gemini or DeepSeek can help with general, public knowledge, but company data should never go into them unless approved and enterprise-protected.
The four rules of safe use
Classify before you share. Is this public, internal, confidential, customer or sensitive information?
Company data goes into Copilot, not public tools. When in doubt, keep it in Copilot.
Never paste customer records, financials, credentials, contracts or personal data into a public tool.
You stay accountable. AI drafts, you decide. Always check the output before you use it.
What can I put into AI?
- General questions and public knowledge, for example "explain 5G in simple terms".
- Text you wrote yourself with no confidential content, for tidying and clarity.
- Practice tasks, learning and skill building.
- Anything already published by MTN and in the public domain.
- Internal notes and drafts with no customer or financial data. Prefer Copilot.
- Meeting notes once names and sensitive details are removed.
- Summaries of internal documents, kept in Copilot with enterprise protection.
- If unsure, anonymise first, then decide.
- Customer names, numbers, account or usage data.
- Financial figures, forecasts, pricing and commercial terms.
- Contracts, legal documents and anything under NDA.
- Passwords, credentials, security details and personal data of employees.
- Anything marked confidential or restricted.
Safe use decision tree
- Ask: what type of information is in this task?
- If it is public or your own general text, any approved tool is fine.
- If it is internal, use Copilot. If it must go elsewhere, anonymise it first.
- If it is customer, financial, confidential or sensitive, use Copilot only, or do not use AI for this part.
- Before you use the output, check it, confirm anything uncertain, and take ownership of the final version.
Classify one real task before you use AI on it
Help me classify this task for safe AI use. Ask me what type of information I am working with, whether it includes internal data, customer data, financial data or confidential information, then recommend whether I should use an enterprise-approved tool, anonymise the information, or not use AI at all.
Pick a task from your week, run the classifier, and note which tool is right.
AI Basics
Understand what AI is and is not, in plain language. Drop the fear, and get ready to try it on one low-risk task.
A simple, honest explanation of AI with no technical overload.
Fear usually comes from the unknown. Understanding the basics turns worry into curiosity.
Read the cards, take the myth quiz, build your context blocks, then improve one email.
The writing coach below. A safe, low-risk first win.
What AI is
Today's workplace AI is a very capable pattern and language tool. It reads your request and predicts a useful response based on huge amounts of text. Think of it as a fast, tireless assistant that drafts, summarises and explains.
What AI is not
It is not conscious, not always correct, and not a source of truth. It has no judgement and no accountability. That stays with you. It can sound confident and still be wrong, so you check its work.
Types of AI, explained simply
Creates text, summaries, images and drafts from a prompt. Copilot and ChatGPT are examples. This is the type you will use most.
Forecasts and spots patterns, for example predicting network demand or customer churn. Works quietly inside many systems.
AI that can take steps to complete a task, not just answer. Early days for most teams. You meet this properly at Level 5.
Why AI matters at MTN
MTN is a large telecoms group with a lot of repeatable, text-heavy work: reports, briefings, summaries, planning, stakeholder prep. AI does not replace the thinking, it removes the friction around it, so people spend more time on judgement and less on the blank page.
AI will not simply take your job. People and teams using AI well will outperform those who do not. The goal here is to get you and your team on the confident side of that line.
Myths and realities
Reality: tasks change, roles adapt. The people who use AI well become more valuable, not less. The risk is not AI, it is standing still while others learn.
Reality: it can invent details, called hallucination. Always check facts, figures and names before you rely on them.
Reality: if you can write an email, you can use AI. The skill is clear instructions, not code.
Reality: it is safe when you follow Level 0. Use Copilot for company data and keep sensitive information out of public tools.
Myth or reality?
"If AI gives me a confident answer with specific figures, I can trust the figures."
Give AI context: your reusable building blocks
AI does its best work when it knows who you are, how you write and what you are working on. Build these short context documents once, then paste the ones you need at the start of any chat. They make AI sound like you, and they keep it honest.
How you want AI to write, so drafts sound like you and not a template.
Use this voice whenever you write for me. Tone: [e.g. clear, warm, professional, direct]. Style: short paragraphs, plain UK English, no jargon, no hype, no long dashes. Vary sentence length so it reads naturally. Words or phrases to avoid: [anything you dislike]. My usual sign-off: [your sign-off]. Acknowledge you have understood, then wait for my task.
Who you are at work, so AI pitches its help at the right level.
Keep my role context in mind when you help me. My role: [your job title]. What I am responsible for: [main responsibilities]. Who I work with: [teams and stakeholders, by role not name]. Decisions I own: [what I decide]. What good looks like in my role: [my quality bar]. Acknowledge, then wait for my task.
What you are working on right now, so answers are relevant, not generic.
Here is my current project context. Project: [name and one-line goal]. Why it matters now: [the driver or deadline]. Key stakeholders: [roles, not confidential names]. Current status: [where things stand]. What I need help with most: [the recurring task]. Do not assume anything not stated here. Ask me if something important is missing.
Add this to any prompt when you want output that does not read like a robot.
Write in plain, human UK English. No hype, no buzzwords, no clichés, no long dashes. Do not open with lines like "In today's fast-paced world". Avoid words such as unlock, leverage, seamless, robust, transformative and delve. Vary sentence length. Sound like a thoughtful colleague, not a brochure.
Add this when accuracy matters, so AI flags what it is unsure of instead of guessing.
Accuracy matters more than confidence. If you are not sure about a fact, figure, name or date, say so clearly rather than guessing. Do not invent sources, statistics or quotes. At the end, list anything I should verify independently before I use this, and flag any assumptions you made.
Want all of these combined into one reusable document you paste at the top of every chat? Open the Context Builder.
Fill in the voice and role blocks, save them, and paste them into your next chat.
Improve one low-risk email or meeting note
Act as a helpful workplace writing coach. Improve the clarity, tone and structure of the text below. Keep the meaning the same. Make it professional, concise and easy to read. Before rewriting, identify three small improvements you will make. [Paste your text here]
Run the writing coach on one real email or note. Notice the three improvements it names.
First Wins
Get immediate value from AI in everyday work: meetings, emails, reports, research and planning. Save time today, safely.
Five-minute missions that turn everyday tasks into quick wins.
The first real time saving is what makes AI stick. Value beats theory.
Pick one mission, copy the prompt, run it in Copilot on real work.
Six ready-to-use prompts, one per common task.
Good prompts read like a good brief to a capable assistant. Say the role, the task, the context, the output you want and any constraints. You practise this in Prompt Better.
Five-minute missions
Each one is copy, paste and go. Tick them off as you complete them.
Act as a workplace writing coach. Rewrite the email below to be clear, warm and professional. Keep it concise. Preserve my intent. Then give me a shorter version I could send if I am busy. [Paste email]
Act as a structured business assistant. Turn the notes below into: 1. Key decisions, 2. Action items with owners, 3. Risks or open questions, 4. Suggested follow-up email. Do not invent missing information. Highlight anything that needs confirmation. [Paste notes]
Act as my meeting prep assistant. I have a meeting about [topic] with [role of attendees]. Help me prepare: 1. The outcome I should aim for, 2. Three key points to make, 3. Likely questions or objections, 4. Two questions I should ask them. Keep it practical.
Act as a business writer. Draft a one-page briefing on [topic] for [audience]. Structure it as: context, what is happening, why it matters, options, and a clear recommendation. Keep it under 300 words and plain-spoken.
Act as a process assistant. Create a clear, practical checklist for [task]. Order the steps logically, flag anything easy to forget, and mark which steps need a second person to check.
Explain [concept] in simple terms for a [audience, e.g. non-technical manager]. Use one everyday analogy. Then give me three sentences I could use to explain it to my team.
Always finish with a hallucination check
Review your previous answer. List any facts, figures, names or claims that I should verify independently before using this. Flag anything you are not confident about, and tell me what you might have assumed.
Make this a habit. Two minutes of checking protects your credibility.
Run Mission 2 on a real set of notes.
Pick any mission above and use it on live work.
Team Workflows
Move from personal productivity to team improvement. Spot where AI could improve a real workflow, and where it should not be used yet.
A way to find and score real AI opportunities in your team's work.
AI on a broken process just breaks things faster. Good workflow thinking comes first.
Map one recurring task, then score it for value, risk and ease.
The workflow improvement coach below.
A tyre-swap booking system used AI to speed up bookings. Customers could book a swap in seconds. But the underlying workflow was never properly designed. The system did not account for the operational reality of retrieving each customer's stored tyres. Bookings succeeded, operations could not keep up, and the result was failed appointments and frustrated customers. The lesson: AI speeds up execution. If the workflow underneath is weak, AI creates bigger problems faster. Fix the process, then add the AI.
Workflow before AI
Before you automate anything, understand the current flow. Where are the inputs, the decisions, the handoffs, the delays, the outputs? Where does judgement matter? Where do errors creep in? AI belongs on the parts that are repetitive and low risk, not the parts that need human judgement or clean data you do not yet have.
Where AI helps, and where it does not, yet
- Summarising long documents and threads
- Drafting first versions of routine communications
- Turning notes into structured outputs
- Comparing options and surfacing questions
- Decisions with legal, financial or safety weight
- Processes that are already broken or undefined
- Anything needing data you cannot share safely
- Work where a wrong answer is hard to catch
Score your use case
Choose one recurring team task and map where AI could support it
Act as a workflow improvement coach. Help me analyse the task below. Ask me questions about the current workflow, inputs, decisions, handoffs, risks, delays and outputs. Then identify where AI could help, where human judgement is still required, and what should be fixed before any AI support is introduced. [Describe the task]
Run the coach on a recurring task. Note one thing to fix before adding AI.
AI Factory Lite
A safe space to test an AI use case before wider adoption. Turn a use case into a small experiment and decide whether to adopt, improve, pause or stop.
A simple ten-step process to test AI ideas safely with real evidence.
Small experiments beat big rollouts. You learn what works before you commit.
Take one use case and design a 30-minute experiment with the prompt below.
The AI experiment designer.
The AI Factory process
- Find the problem
- Describe the current workflow
- Identify where AI might help
- Write a first prompt
- Test with safe data
- Check the output
- Gather feedback
- Improve the prompt or workflow
- Assess risk and value
- Decide: adopt, pause, escalate or stop
Name the real problem and turn it into a clear instruction AI can act on.
Run the prompt on safe test data and produce a first working output.
Put it in front of a real user, gather feedback, check the output honestly.
Experiment scorecard
Always test with safe or anonymised data inside Copilot. An experiment is not a reason to relax the Level 0 rules. Build your evidence pack as you go: the problem, the prompt, the result, the feedback and the recommendation.
Take one use case and design a 30-minute AI experiment
Act as an AI experiment designer. Help me turn this use case into a small, safe test. Create: 1. Problem statement, 2. Target user, 3. Current workflow, 4. AI-supported workflow, 5. Test prompt, 6. Safe test data required, 7. Success criteria, 8. Risks, 9. Feedback questions, 10. Adoption recommendation template. [Describe your use case]
Produce a plan, run it on safe data, and write a one-line recommendation.
AI Champions
Support others, share responsible wins, spot future automation, and understand when more advanced solutions are appropriate, and when they are not.
The path from confident user to the person who helps a whole team.
Adoption spreads person to person. Champions turn one win into many.
Design a 20-minute team practice session with the prompt below.
The AI adoption coach.
- Help colleagues get their first safe win
- Share wins openly, without hype
- Model safe use every time
- Point people to Copilot and the Level 0 rules
- Reusable prompt packs for common team tasks
- Team context documents so AI knows your world
- Simple adoption stories others can copy
- A shared place to collect what works
Introduction to agentic workflows
Agentic AI can take several steps to complete a task, not just answer a question. It might read a brief, draft a document, and prepare a summary in sequence. Powerful, but it needs clear guardrails, good data and human checkpoints. Treat it as the next frontier, approached carefully.
If the process is unclear, the data is sensitive, the stakes are high, or a wrong step is hard to catch, keep a human in control. Automation amplifies whatever it is built on, including the flaws.
Governance handoff checklist
- Is the use case documented with an evidence pack?
- Has data handling been checked against Responsible AI and Level 0?
- Are the risks named, with an owner?
- Is there a human accountable for outputs?
- Has it been shared with the right governance contact before scaling?
Design a 20-minute AI practice session for your team
Act as an AI adoption coach. Help me design a 20-minute practice session for my team. The goal is to help them get one safe first win using AI. Include: opening message, safety reminder, task, copy-paste prompt, reflection question, and suggested next step.
Adoption story template
Help me write a short adoption story to share with colleagues. Structure: the task that used to take too long, the safe prompt I used, the result, the time saved, and one tip for someone trying it. Keep it under 150 words, plain-spoken, no hype.
Design and deliver a 20-minute session that gets one colleague their first win.
Tool Safety Selector
Not sure which tool is safe for your task? Choose the type of information, and this tells you what to do. When in doubt, Copilot.
Step 1 — What kind of information is in this task?
Copilot vs public AI tools
Copilot — MTN standard
Company data stays inside MTN's protected environment. Use for anything touching MTN information.
Public tools
ChatGPT, Perplexity, Claude, Gemini, DeepSeek. Useful for public knowledge. Never enter company, customer, financial or confidential data unless approved with enterprise protection.
MTN may consider other tools in future. Until then, Copilot is the standard for work involving MTN data.
Prompt Better
A good prompt is a good brief. Learn the six-part pattern that turns vague requests into reliable results.
Prompt like a manager: the six-part pattern
"Write something about our new plan."
No role, no audience, no format. You get generic filler.
"Act as a business writer. Draft a 200-word update on our new plan for team leads. Plain-spoken, three short paragraphs, end with one clear action."
Two moves that instantly improve results
Before you answer, ask me up to five questions that would help you give me a better result. Wait for my answers, then complete the task.
Act as a constructive critic. Review my plan below. Point out the three weakest assumptions, what I might be missing, and one risk I have not mentioned. Be direct but practical. [Paste your plan]
Common prompting mistakes
- Being vague. If you would not accept it from a colleague, do not accept it in a prompt.
- No context. AI cannot read your mind or your files unless you tell it.
- Asking for too much at once. Break big tasks into steps.
- Trusting the first draft. Iterate: "make it shorter", "more formal", "add an example".
- Skipping the fact check. Confident does not mean correct.
Context Builder
AI gets better when it knows your world. Build a personal context document once, reuse it every day. Paste it at the start of a chat and AI works the way you do.
What goes into a personal context document
- Role context — your job, what you are responsible for
- Team context — who you work with, your function's goals
- Writing style — plain, warm, concise, UK English
- Quality criteria — what a good output looks like to you
- Anti-AI-writing rules — no hype, no clichés, no filler
- Hallucination checks — always flag facts to verify
Your reusable context prompt
Fill in the brackets once, save it, and paste it at the top of any chat.
From now on, use this context when helping me. My role: [your role at MTN and main responsibilities] My team: [your function, who you work with, current priorities] My audience: [who I usually write for] My writing style: clear, warm, concise, plain-spoken, UK English. No hype, no clichés, no jargon, no long dashes. Short paragraphs. What good looks like: [your quality standard] Always: flag any facts, figures or names I should verify. Ask me a question if you are missing something important. Do not invent details. Acknowledge you have understood, then wait for my task.
Anti-AI-writing instructions
Want output that does not sound like a robot? Add this line to any prompt.
Write in plain, human, UK English. No hype, no buzzwords, no clichés, no long dashes. Vary sentence length. Sound like a thoughtful person, not a marketing brochure.
Fill in the reusable prompt, save it somewhere handy, and use it in your next chat.
Prompt Library
Prompts organised by the job you need done. Each includes what it is best for, a safety reminder, the copy-paste prompt, an example of the output, and how to improve it.
Use Case Spotter
Where is AI most useful in your role? Look for repeatable tasks, time drains and blank-page work, then score each idea for value, risk and ease.
A quick way to find real AI opportunities in your own work.
The best use cases are hiding in your weekly routine, not the headlines.
List three tasks, score them, pick the highest value and lowest risk.
The use case finder below.
Where to look
- Repeatable tasks — anything you do the same way each week
- Time drains — work that takes longer than it should
- Blank-page work — drafting from scratch
- Summarise, draft, analyse — the three AI is best at today
Score each use case (1 to 5)
High value, low risk, high ease. That is your first experiment. Take it to AI Factory Lite.
Find and score three use cases in your role
Act as an AI opportunity spotter. Ask me about my role, my weekly tasks, and where I lose the most time. Then suggest five specific tasks where AI could help. For each, give a value, risk and ease rating out of 5, and recommend which one to try first.
Run the finder, score your ideas, and pick one to submit.
Manager Pathway
Your team takes its lead from you. This is how to reduce fear, encourage safe play, and build AI habits without turning it into surveillance.
- Name the fear openly. Most people worry AI will take their job.
- Reframe it honestly: those who use AI well will do better than those who do not.
- Go first. Share your own clumsy early attempts.
- Give people permission to experiment on low-risk tasks.
- Point everyone to Level 0 and Copilot first.
- Make the first win small and achievable.
- Run a short AI practice session, monthly.
- Keep a shared prompt pack for common team tasks.
- Celebrate wins, however small.
- Do not track usage to judge people.
- Frame AI as support, not monitoring.
- Focus on outcomes and confidence, not compliance.
Collect and share first wins
Adoption spreads through stories, not mandates. When someone on your team saves an hour with a good prompt, capture it and share it. One real example does more than any policy.
Help me write three short questions I can ask my team each month to collect their AI wins. The questions should surface what task they used AI for, the prompt that worked, and the time or effort saved. Keep it light and encouraging.
Identify team use cases together
Act as a facilitator. Design a 30-minute team session to identify AI use cases. Include: an opening reframe to reduce fear, a safety reminder, a simple exercise to list repeatable and time-consuming tasks, a way to score them by value, risk and ease, and a close that picks one use case to test.
Use the Level 5 session designer or the workshop prompt above.
Confidence Score
Rate yourself honestly on each statement from 1 to 5. Do it now, before you begin, and again after a few weeks. Watch the number move.
Rate all eight statements to see your band and a suggested starting level.
A low score is not a problem, it is a starting point. Anything under 3 on a statement points you to the level that builds it. Come back in a few weeks and re-rate.
Next Steps
You do not need to do everything. Do the next thing. Here is the order that works.
- Complete the Responsible AI prerequisite. Level 0 is your practical companion to it.
- Try three first-win prompts. From Level 2, on real work, this week.
- Build your personal context document. Use the Context Builder. Reuse it every day.
- Submit one use case. Find it in the Use Case Spotter, shape it in AI Factory Lite.
- Join a masterclass. The academy is your on-ramp before, during and after the live sessions.
- Join an AI Factory sprint. Turn your use case into a tested experiment.
- Become an AI Champion. Help your team get their first wins.
Finish Level 0 and get one first win. That is it.
Build your context doc and submit one use case.
Run an experiment and share a win with your team.
AI will not simply take your job. People and teams who use it well will outperform those who do not. You are already on the right side of that line. Keep going.