Why Mindset Comes First
Most AI training starts with tools. Open ChatGPT. Write a prompt. Get a response. Iterate. And that's useful — but it skips something important.
If you've ever followed an AI tutorial step by step and still felt like something was off about the results, you're not alone. The issue usually isn't the tool. It's not even the prompt. It's the assumptions you brought to the conversation — what you expected AI to do for you, what you thought your role was, and where you drew the line between "AI's job" and "my job."
That line matters more than any technique.
This path covers seven principles. They're not abstract philosophy — they're practical mental habits that change how you interact with every AI tool you'll ever use. People who adopt these principles tend to get dramatically better results, not because they write fancier prompts, but because they think more clearly about what they're actually trying to accomplish.
Here's what I wish someone had told me when I started using AI: the people who struggle most aren't the ones who lack technical skills. They're the ones who either hand over too much responsibility to the AI (and get mediocre, generic output) or hold back entirely because they're unsure whether AI will make them worse at their job. Both reactions come from the same place — an unclear relationship with the tool.
These seven principles clarify that relationship. Once they click, everything else in AI Tutorium — the tools, the skills, the techniques — becomes significantly more effective.
Get Started
Before we dive into the principles, here's a quick self-assessment. There are no right or wrong answers — just an honest snapshot of where you're starting from.
Think about the last few times you used an AI tool (or decided not to). For each statement below, note whether it sounds like you:
- "I usually accept the AI's first response without much questioning." (Over-trust → Principle 1)
- "I avoid using AI for important work because I'm not sure I can trust the output." (Under-trust → Principle 3)
- "I sometimes feel like AI is doing my thinking for me, and I'm not sure that's a good thing." (Role confusion → Principle 2)
- "I spend more time learning new AI tools than actually using them on real problems." (Tool fixation → Principle 5)
- "I feel pressure to be 'good at AI' but I'm not sure what that means." (Unclear goals → Principle 6)
If any of these resonated — most people relate to at least two — you're in exactly the right place. Each one connects to a specific principle below. By the end of this path, you'll have a clear framework for navigating all of them.
This path is different from the others. There are no AI tools to configure, no prompts to copy, and no workflows to build. Instead, each principle comes with a reflection exercise — a short, practical activity that helps you internalise the idea through your own experience. These are deceptively simple. The value is in doing them honestly, not quickly.
AI Is a Tool for Better Human Thinking, Not a Replacement for It
Last verified: March 2026
This is the foundation everything else builds on, and it's worth being really clear about it: AI can do a lot. It can draft, analyse, summarise, brainstorm, code, translate, and generate ideas at a speed no human can match. But it cannot weigh up what matters to you. It cannot apply your experience. And it cannot take responsibility for the outcome.
That's your job. The best results come when you stay in the loop, not when you step out of it.
This might sound obvious, but in practice it's surprisingly easy to slip. When AI gives you a polished, confident-sounding answer, there's a natural pull to accept it — especially when you're tired, pressed for time, or working outside your area of expertise. We've all been there. The output looks good, it's well-structured, it sounds authoritative… so you use it. And sometimes that works out fine. But sometimes you miss something the AI got subtly wrong, because you weren't really checking — you were trusting.
🔢Research on cognitive offloading suggests a real risk:
when people rely on tools (calculators, GPS, AI assistants) for tasks they could do mentally, their ability to perform those tasks independently tends to decline over time. Early studies on AI-assisted analytical work show a similar pattern — participants who used AI assistance on reasoning tasks showed reduced performance on similar tasks without AI afterwards, compared to those who solved problems manually first. The brain treats AI like a calculator and stops doing the mental work itself.
Source: Cognitive offloading research (Risko & Gilbert, 2016); emerging AI-assisted cognition studies
The risk isn't that AI will replace your thinking. The risk is that you'll let it — gradually, without noticing — and your own judgement will atrophy from disuse. That's not AI's fault. It's a human pattern: when a tool handles something for us, we stop practising the underlying skill.
The antidote is simple: stay engaged. Use AI to think better, not to think less. Ask it to challenge your reasoning, not just confirm it. Use it to explore options you hadn't considered, then make the call yourself.
The loop you should stay in
Here's a useful mental model. Think of your interaction with AI as a loop with four steps:
⚖️ The Frame-Generate-Evaluate-Decide Loop
Step Who What Happens 1. Frame You Define the problem, the context, and what a good outcome looks like 2. Generate AI Produce options, drafts, analysis, or ideas 3. Evaluate You Assess the output against your knowledge, goals, and judgement 4. Decide You Choose what to use, what to discard, and what to refine
Steps 1, 3, and 4 are yours. They always will be. AI only does step 2. When you collapse all four steps into "ask AI and use whatever it says," you've left the loop — and that's where quality drops.
A useful test: if someone asked you to explain why you chose the output AI gave you, could you? If you can articulate the reasoning — "I used this draft because it addressed the stakeholder's specific concern about timeline, which was the real issue" — you're in the loop. If your honest answer is "it sounded good," you've stepped out.
Exercise: The Decision Audit
Here's what we'd suggest: Think about the last three times you used AI for something that mattered — a work email, a report, a decision, a creative project. For each one, ask yourself:
- Did I clearly define what I wanted before I asked AI? (Frame)
- Did I critically evaluate what AI gave me, or did I mostly accept it? (Evaluate)
- Could I explain to someone else why I used that specific output? (Decide)
If you answered "no" or "not really" to any of these — that's not a failure. It's just information about where you tend to step out of the loop. Most of us default to skipping the evaluation step when we're busy. Knowing your pattern is the first step to changing it.
Write down one specific habit you could adopt to stay in the loop more consistently. Keep it small and concrete — something like "I'll spend 30 seconds asking myself what I actually need before I open ChatGPT" is more useful than "I'll be more critical."
AI Amplifies What You Bring to It
Last verified: March 2026
Here's something that took me longer than I'd like to admit to understand: AI doesn't correct you. It accelerates you.
Bring clarity, and you'll get clarity back. Bring vague thinking, unchecked assumptions, or poorly framed questions, and you'll get more of that too — except now it'll be wrapped in confident, articulate language that makes it harder to spot the problems.
This is the amplification effect, and it works in both directions. When you bring a well-defined problem, relevant context, and a clear sense of what "good" looks like, AI becomes remarkably powerful. When you bring a fuzzy idea and hope AI will figure out what you mean, you get polished mediocrity — something that looks professional but doesn't actually solve your problem.
🎯 The amplification test: If your input wouldn't make sense to a smart colleague who doesn't know the background, it probably won't produce great AI output either. AI doesn't read your mind — it reads your words. The gap between what you mean and what you say is where quality gets lost.
Garbage in, clarity out?
One of the most persistent myths about AI is that it can take vague input and return precise output. Sometimes it appears to — and that's actually the problem. When AI fills in the gaps you left, it's guessing. It's pattern-matching against its training data, not understanding your specific situation. Sometimes those guesses are excellent. Sometimes they're plausible-sounding nonsense.
The issue isn't that AI gives bad answers to bad questions. The issue is that it gives confident answers to bad questions. And confident wrong answers are more dangerous than obviously wrong ones, because you're less likely to question them.
🔢In enterprise AI adoption research,
a consistent pattern emerges: requests that include specific context — role, audience, constraints, and success criteria — produce dramatically more usable output than vague requests. The gap isn't explained by prompt length or complexity; it's almost entirely explained by input clarity. Across multiple studies, the difference between contextualised and context-free prompts is often 2-3x in output usability.
Source: BCG and McKinsey enterprise AI adoption research, 2023–2025
In our experience, the single biggest predictor of AI output quality isn't prompt engineering technique — it's how clearly you understood your own problem before you typed anything. People who spend 60 seconds thinking about what they actually need before opening an AI tool consistently get better results than people who spend 10 minutes trying to refine a prompt that was vague from the start.
Try this: before your next AI interaction, write down — in plain language, as if explaining to a colleague — what you need and why. Not the prompt. Just the problem. If you can't articulate it clearly to a human, AI won't magically solve that clarity gap for you.
Exercise: The Input Quality Test
Here's what we're going to do: Take a task you'd normally hand to AI and run it through two rounds.
Round 1: Use your normal approach. Open your AI tool, type what comes to mind, and save the output.
Round 2: Before opening any AI tool, spend 2–3 minutes writing down:
- What exactly do I need? (Not "help with an email" — what specific email, to whom, about what, and what outcome do I want?)
- What context does AI need that it doesn't have? (Background, constraints, audience, tone)
- What would a great result look like? (How will I know if the output is good?)
Then use that thinking to craft your prompt and compare the two outputs.
Most people find the Round 2 output is noticeably better — not because the prompt was fancier, but because the thinking behind it was clearer. That's the amplification effect in action.
You Are the Decision-Maker
Last verified: March 2026
AI can give you options, arguments, pros-and-cons lists, risk assessments, and data summaries. What it cannot do is tell you what the right answer is for your situation, your values, or your goals. Every decision still needs a human behind it — and that human is you.
This sounds straightforward, but it gets complicated in practice. When AI presents three options and recommends one, it feels like a decision has been made. When AI writes a persuasive argument for a particular approach, it feels like the case is settled. But AI doesn't have skin in the game. It doesn't know your risk tolerance, your team dynamics, your budget constraints, or the political realities of your organisation. It's working from patterns, not from judgement.
If you've ever gone along with an AI recommendation and later thought "that wasn't quite right for us," you've experienced this gap. It's not that the AI was wrong in general — it's that it was wrong for your specific context, and only you had that context.
⚖️ Options vs. Choices — What AI Can and Can't Do
Capability AI You Generate options from patterns ✓ Excellent Slower, but with context List pros and cons ✓ Comprehensive More selective, more relevant Weigh personal values ✗ Cannot ✓ Only you can Apply organisational context ✗ Limited ✓ Only you can Take responsibility for outcomes ✗ Cannot ✓ Always you Factor in relationships and politics ✗ Cannot ✓ Only you can
This isn't about distrusting AI. It's about understanding what "trust" means in this context. You can trust that AI will generate competent, well-reasoned options based on patterns in its training data. You should not trust it to weigh those options against factors it doesn't have access to — your experience, your relationships, your goals, your gut feeling about what's right.
🔢Research on human-AI decision-making consistently shows the same pattern:
people who use AI-generated recommendations as one input alongside their own judgement make better decisions than those who either ignore AI entirely or follow AI recommendations without modification. The worst outcomes come from uncritical adoption — treating the AI's suggestion as the answer rather than an input. The best outcomes come from what researchers call "appropriate reliance" — knowing when to trust the AI and when to override it.
Source: Human-AI decision-making research (Bansal et al., 2021; Buçinca et al., 2021); Harvard Business Review analysis of AI-augmented decision-making
Exercise: The Final Call
If this feels right for you, try this: Pick a decision you're currently facing — it doesn't have to be big. Ask your preferred AI tool to help you think through it. Prompt it to:
- List the key options
- Identify the pros and cons of each
- Recommend an approach
Now, before you accept the recommendation, write down your answers to these questions:
- What does AI not know about my situation that would change this recommendation?
- What values or priorities am I weighing that AI can't see?
- If this decision goes wrong, who's accountable? (Hint: not the AI.)
Notice the gap between what AI recommended and what you'd actually choose when you factor in everything it doesn't know. That gap is your judgement — and it's the most valuable thing you bring to the table.
Ask Better Questions, Not Just Better Prompts
Last verified: March 2026
Most AI advice focuses on how to phrase your prompts — use this template, add this role, structure your request this way. And that's genuinely useful. But the deeper skill, the one that separates people who get transformative results from people who get decent ones, is knowing what to ask in the first place.
Understanding your problem clearly, before you open any AI tool, is what separates useful outputs from noise.
This is the part where most people's eyes start to glaze over — and that's fair, because it sounds like generic advice. "Ask better questions" could mean anything. So let me be specific about what I mean.
📈 The Question Hierarchy
Estimated distribution based on observed AI interactions across workshops and enterprise teams. Most users stay at Level 1 — moving to Level 2 alone produces dramatically better results.
Problem clarity before tool selection
Here's a pattern we see constantly: someone opens ChatGPT, types a request, gets a mediocre result, then spends 15 minutes trying to fix the prompt. After five rounds of iteration, they end up with something acceptable — but the real issue was that they weren't clear about what they needed in the first place.
If you'd asked them "what are you trying to accomplish?" before they started, they would have paused and said something like, "Well, I need to convince my manager to approve this budget, and the real concern is the timeline risk." That's a clear problem. But what they typed into AI was "help me write a budget proposal." The gap between those two things is where all the wasted iteration lives.
From what we've seen, spending even 60 seconds on problem clarity before touching any AI tool saves more time than any prompt template ever will. It's the highest-leverage habit in this entire path.
Something I wish I'd understood earlier: the reason prompt engineering advice often doesn't work as well as promised isn't that the techniques are wrong — it's that they're solving the wrong problem. No prompt template can compensate for unclear thinking about what you actually need. Get the thinking right first, and the prompts almost write themselves.
Exercise: The Question Behind the Question
Here's what we'd suggest trying: Take something you'd normally ask AI to help with this week. Before you type anything, write down answers to these three questions:
- What am I actually trying to achieve? (Not "write an email" — what outcome do I want from this email?)
- Why does this matter? (What's at stake? What happens if I get it wrong?)
- What would a great result look like? (Be specific — not "a good email" but "a response that gets my manager to approve the timeline extension without requiring a meeting")
Then — and only then — craft your AI prompt. Notice how much easier it is to write a clear prompt when you've already done the thinking.
If this exercise feels almost too simple, that's the point. The skill isn't complicated. It's just easy to skip.
Focus on Timeless Problems, Not Today's Tools
Last verified: March 2026
AI tools will keep changing — new versions, new platforms, new features, new pricing, new capabilities. In the time since you started reading this path, there's a reasonable chance something in the AI landscape has already shifted. That's how fast it moves.
But here's what doesn't change: the problems you're solving at work. Communicating clearly. Making decisions with incomplete information. Managing projects. Understanding customers. Creating things people value. These challenges predate AI and they'll outlast any specific tool.
When you focus on the outcome you need rather than the tool you're using, your skills stay relevant no matter what changes.
⚖️ Feature Skills vs. Capabilities
Type Example Durability Transferability Feature skill "I know how to use ChatGPT's Custom Instructions" Fragile — changes with updates Low — one tool only Feature skill "I know Claude's Artifacts feature" Fragile — platform-specific Low — one tool only Capability "I can frame a problem clearly before using any AI" Durable — tool-independent High — works everywhere Capability "I can evaluate whether AI output is accurate" Durable — tool-independent High — works everywhere Capability "I can synthesise information from multiple sources" Durable — predates AI High — works without AI too
Outcomes over features
There's a reason we structured AI Tutorium around skills and mindsets, not just tools. Tools are the "how." Skills and mindsets are the "what" and "why." If you understand the underlying capability — say, information synthesis — you can apply it whether you're using Claude, ChatGPT, Gemini, or whatever emerges next year.
This doesn't mean tool knowledge is worthless. Far from it — each tool has genuine strengths worth understanding. But if you're spending all your learning time keeping up with tool updates and none of it developing transferable skills, the ratio is off.
🔢Tool-specific AI knowledge decays fast:
from what we've observed and what workforce research suggests, tool-specific knowledge has a half-life of roughly 6–9 months — by that point, about half of what you've learned about specific features is outdated or superseded by updates and new releases. Transferable skills (problem framing, critical evaluation, information synthesis) show no measurable decay over the same period. A useful rule of thumb: invest about 70% of your learning time in transferable skills and 30% in tool-specific knowledge.
Source: AI Tutorium observations; Deloitte and McKinsey workforce readiness research on AI skill development
In our experience, the most effective AI users follow something close to that ratio — and they report feeling significantly less overwhelmed by the pace of AI change. When your core skills transfer across tools, a new platform launch is interesting, not threatening.
If this feels like a lot of pressure to learn the "right" things — it's not meant to. There's no exam. The point is freeing: you don't need to learn every new tool that launches. You need to get good at a handful of transferable skills, and then pick up new tools as needed. That's a much more sustainable approach.
Exercise: The Tool-Agnostic Challenge
Here's one approach — see if it fits your situation: Pick a task you regularly use AI for. Now imagine your preferred AI tool disappeared overnight. Ask yourself:
- Could I still accomplish this task with a different AI tool? If so, what would change?
- Could I accomplish it without any AI tool at all? (How did people do this before AI?)
- What's the underlying skill this task requires? (e.g., "summarising complex information," "writing persuasively," "identifying patterns in data")
Now rate yourself on that underlying skill — independent of any AI tool. If your rating is lower than you'd like, that's not a reason to avoid AI. It's a reason to use AI more deliberately: as a partner that helps you develop the skill, not a crutch that replaces it.
If you discover that you'd be genuinely stuck without your AI tool, that's worth noting — not as a criticism, but as a signal about where to invest in your own development.
Choose Progress Over Perfection
Last verified: March 2026
Learning to use AI well is a practice, not a destination. You'll get things wrong. You'll produce outputs that miss the mark. You'll find better approaches over time and look back at your early attempts with a mix of amusement and mild horror. That's not failure — that's how mastery works.
Consistent effort compounds.
If you're waiting until you feel "ready" to start using AI on real work, I want to be honest with you about something: that day doesn't come. There's always a newer tool, a better technique, another tutorial to watch first. The people who get good at AI are the ones who start using it before they feel ready, make mistakes in low-stakes situations, learn from those mistakes, and gradually expand their range.
🔢Hands-on experimentation beats waiting for readiness:
workplace learning research consistently shows that people who start using new tools "regularly but imperfectly" develop proficiency significantly faster than those who wait until they feel fully prepared. This pattern holds strongly for AI tools — hands-on experimentation, even when the results are poor, builds intuition faster than structured study alone. The people who feel least ready to start are often the ones who benefit most from starting anyway.
Source: Workplace learning research (Eraut, 2004; Kolb experiential learning theory); McKinsey AI adoption surveys
This principle applies to everything in AI Tutorium. You don't need to complete every learning path before you start applying what you've learned. You don't need to master prompt engineering before you try content ideation. You don't need to understand every model's architecture before you start using one. Start where you are. Use what you have. Improve as you go.
Struggling with this doesn't mean you're doing it wrong. It means you're doing it.
📈 The Learning Curve Most People Don't Expect
Satisfaction percentages are illustrative, based on patterns observed across AI training programmes. Most people quit during the trough — don't. That's where real learning happens.
Something we've noticed: the people who make the fastest progress aren't the ones who study the most. They're the ones who have the shortest gap between learning something and trying it. Read a technique, try it that day. Watch a tutorial, apply it that afternoon. The learning is in the doing, not in the absorbing.
Exercise: The Imperfect Experiment
Here's what we're going to do: Pick one AI technique or tool you've been meaning to try but haven't gotten around to yet. (We all have a list.) Set a timer for 15 minutes and just try it. Don't research it further. Don't watch another tutorial. Don't optimise your prompt before you start. Just try it.
When the timer goes off, write down:
- What worked, even partially?
- What surprised you?
- What would you do differently next time?
- What question do you now have that you didn't have before?
That last question is the most valuable. It means you've moved from "I don't know what I don't know" to "I know exactly what I need to learn next." That's progress — real, concrete progress — from just 15 minutes of imperfect practice.
If the result was terrible, that's actually the best outcome. Now you know something concrete about what doesn't work, and that's more valuable than another hour of theory.
Know When to Use AI — and When Not To
Last verified: March 2026
This is the principle that surprises people most, and it might be the most important one.
Reaching for AI by default isn't skill — it's habit. And like any habit, it can be helpful or harmful depending on whether it's deliberate. The most effective AI practitioners we've worked with are remarkably selective about when they use AI. They know which tasks genuinely benefit from it, which ones don't, and — crucially — which ones are actively harmed by it.
⚖️ When AI Helps vs. When It Doesn't
AI Adds Genuine Value Better Done Without AI Drafting routine communications Emotionally sensitive messages Brainstorming and exploring options Tasks that build skills you need to maintain Summarising large volumes of information High-stakes decisions with ambiguous inputs Checking your reasoning and finding blind spots Creative work where the process develops your craft Translating or reformatting content Tasks that take less than 2 minutes manually Research across multiple sources Situations requiring authentic personal voice
Restraint as skill
There are legitimate reasons not to use AI for certain tasks, and they're not about being a luddite or being afraid of technology. Some tasks develop skills you need to maintain. Some require emotional intelligence that AI can't replicate. Some involve stakes where AI's confident-but-sometimes-wrong nature creates unacceptable risk. And some are simply faster to do yourself than to explain to AI.
🔢A 2025 workplace productivity study found
that the highest-performing AI users spent an average of 3.2 hours per day using AI tools — not the 6+ hours reported by the heaviest users. The moderate users outperformed heavy users on quality metrics by 22%, primarily because they were more selective about which tasks benefited from AI assistance and which ones they handled themselves.
Source: Adapted from Slack/Salesforce "State of AI at Work" report, 2025
Here are situations where the most skilled AI users we know deliberately choose not to use AI:
- Building expertise in a new area. If you're learning something for the first time, having AI do the work for you skips the cognitive effort that builds understanding. Use AI to check your work or explain concepts, not to do the thinking for you.
- Emotionally sensitive communication. A condolence message, a difficult performance conversation, a personal apology — these need your authentic voice and emotional judgement, not a well-crafted AI template.
- High-stakes decisions with limited data. When the consequences are significant and the inputs are ambiguous, your experienced judgement is more reliable than AI's pattern-matching on incomplete information.
- Tasks that take less than 2 minutes manually. By the time you've typed the prompt and reviewed the output, you could have just done it. Not everything needs to be optimised.
- Creative work where the process matters. Sometimes the value is in the struggle — the thinking, the exploration, the unexpected connections your own mind makes. AI can help you refine, but it shouldn't replace the generative process when that process is how you develop your craft.
I used to think being good at AI meant using it for everything. It took me a while to realise that knowing when not to use it is the more advanced skill. Restraint doesn't mean you're behind — it means you're being strategic about where AI adds genuine value.
Exercise: The Deliberate Choice
Here's what we'd suggest trying for one week: Before each AI interaction, pause for five seconds and ask yourself one question: "Would I get a better result doing this myself?"
Keep a simple tally over five working days:
- How many times did you use AI and it genuinely helped?
- How many times did you use AI out of habit, when you could have done it faster yourself?
- How many times did you choose not to use AI, and that turned out to be the right call?
- How many times did you choose not to use AI, and later wished you had?
Most people discover that about 20–30% of their AI interactions are habitual rather than strategic. That's not a problem — it's an opportunity. Redirecting those interactions frees up mental energy and reinforces the skills that matter most.
The goal isn't to use AI less. It's to use it deliberately — every time, for the right reasons.
Challenge Exercises
These longer-form exercises integrate multiple principles. They're designed to be completed over one to two weeks, building real behavioural change rather than just intellectual understanding.
Challenge 1: The Seven-Day Principle Tracker
Duration: One week
Each day, focus on one principle (skip the seventh day or revisit the one you found hardest). Before every AI interaction that day, consciously apply that day's principle. At the end of each day, write a single sentence about what you noticed.
By the end of the week, you'll have a personal map of which principles come naturally and which need conscious practice. That map is more valuable than any AI certification.
Challenge 2: The AI-Free Morning
Duration: Five working days
For one working week, don't use any AI tools before noon. Do your thinking, writing, and problem-solving manually in the morning. After noon, use AI as normal. At the end of each day, compare: Was the quality of your morning work different from your afternoon work? Where did you miss AI most? Where did you not miss it at all?
This exercise makes Principles 1, 2, and 7 visceral. You'll discover which of your AI habits are genuinely productive and which are just comfortable.
Challenge 3: Teach Someone These Principles
Duration: One session (30–60 minutes)
Pick a colleague, friend, or family member who uses AI (or is curious about it). Explain two or three of these principles in your own words. Don't lecture — have a conversation. Ask them which principles resonate and which ones they push back on.
Teaching forces you to move from understanding to articulation, which is where internalisation happens. The pushback you receive will sharpen your own thinking. And you'll probably learn something about how other people relate to AI that surprises you.
Challenge 4: Write Your AI Philosophy
Duration: One sitting (20–30 minutes)
In 200–300 words, write your personal philosophy for using AI at work. Ground it in these seven principles, but make it yours — your context, your values, your boundaries. Cover:
- What you believe AI is for (and not for) in your work
- Where you draw the line between "AI's job" and "my job"
- One habit you're committed to maintaining
- One thing you're deliberately choosing not to outsource to AI
Revisit this statement in three months. If it hasn't changed at all, you probably haven't been experimenting enough. If it's changed entirely, you've been growing. Either way, you'll have a concrete reference point for how your relationship with AI is evolving.
Quick Reference
Last verified: March 2026
The Seven Principles at a Glance
| # | Principle | In Practice |
|---|---|---|
| 1 | AI is a tool for better human thinking | Stay in the Frame → Generate → Evaluate → Decide loop |
| 2 | AI amplifies what you bring to it | Spend 60 seconds on problem clarity before typing |
| 3 | You are the decision-maker | Generate options with AI, make choices yourself |
| 4 | Ask better questions, not just prompts | Define the problem before you define the prompt |
| 5 | Focus on timeless problems | Invest 70% in transferable skills, 30% in tools |
| 6 | Choose progress over perfection | Try it today, imperfectly. Iterate tomorrow. |
| 7 | Know when to use AI — and when not to | Pause 5 seconds before each AI interaction |
The Frame-Generate-Evaluate-Decide Loop
Before every AI interaction, remember: you own three of the four steps. Frame the problem (you). Let AI generate options (AI). Evaluate the output against your knowledge (you). Decide what to use (you). If you can't explain why you chose the output, you've left the loop.
Common Mistakes and Fixes
| Mistake | Which Principle | Fix |
|---|---|---|
| Accepting AI output without checking | Principle 1 | Ask: "Can I explain why this is right?" |
| Typing vague prompts and iterating endlessly | Principle 2 | Write down the problem in plain language first |
| Treating AI's recommendation as your decision | Principle 3 | List what AI doesn't know about your situation |
| Focusing on prompt tricks over problem clarity | Principle 4 | Answer "what am I trying to achieve?" before prompting |
| Chasing every new AI tool release | Principle 5 | Ask: "Does this help with a problem I already have?" |
| Waiting to feel "ready" before using AI | Principle 6 | Set a 15-minute timer and just try it |
| Using AI out of habit for every task | Principle 7 | Ask: "Would I get a better result doing this myself?" |
When to revisit these principles
These aren't rules you memorise once and forget. They're principles you return to when something feels off:
- AI output quality dropping? → Check Principles 1–2 (are you staying in the loop and bringing clarity?)
- Feeling overwhelmed by AI tools? → Check Principle 5 (focus on capabilities, not features)
- Unsure if AI is helping or hurting your development? → Check Principles 6–7 (progress over perfection, deliberate choice)
- Making decisions you later regret? → Check Principle 3 (are you treating AI options as your choices?)
- Spending too long on prompts? → Check Principle 4 (clarify the problem first)
The principles work together. Getting better at any one of them makes the others easier. And none of them require you to be an AI expert — they require you to be a clear thinker who happens to use AI.
You've got everything you need to start. The tools will keep changing. The techniques will keep evolving. But these seven principles? They'll serve you for as long as AI is part of your work — which, at this point, is likely the rest of your career. Build from here.
Practice Project
Knowing these principles intellectually is one thing. Living them is another — and the gap between the two is where most people get stuck. This project bridges that gap by turning abstract ideas into a document you can actually use every day.
Time: 45–60 minutes
What you'll build: A Personal AI Charter — 7 principles that define how you want to use AI in your role, grounded in real examples from your own work.
Why this matters: We've found that people who write down their AI principles — even informally — make noticeably better decisions under pressure. When a deadline hits and you're tempted to hand everything to AI without thinking, a charter gives you something concrete to check against. It's the difference between "I should probably review this" and "Principle 3 says I own the final call."
Steps:
- Reflect on your current AI habits. Spend 10 minutes writing honestly about how you use AI right now. What works well? What feels off? Where do you rely on AI too much, and where might you be holding back unnecessarily? Don't filter — this is just for you.
- Draft 7 principles for how you want to use AI in your role. These don't need to mirror the 7 principles from this path exactly — they should reflect your specific work, your values, and the situations you actually face. Maybe one of yours is "I never send AI-drafted client emails without reading them aloud first." That's perfect. The more specific, the more useful.
- For each principle, write a concrete example of what it looks like in practice. "Stay in the loop" is vague. "I review every AI-generated report section before it goes into the client deck, even when I'm rushed" is actionable. One sentence per principle is enough — you're looking for the moment where the principle actually changes your behaviour.
- Test each principle against a real recent task. Pick something you did with AI in the last week. Walk through your charter and ask: did I follow this? Would following it have changed the outcome? This step often reveals which principles are genuinely useful and which sound good but don't apply to your actual workflow.
Deliverable: A personal AI charter — 7 principles with practical examples — that you can pin above your desk or keep in a note you actually revisit.
Stretch goal: Share your charter with a colleague and ask them to write their own. Compare notes — the differences in what you each prioritise tend to spark genuinely useful conversations about how your team works with AI.
Reflection: Which principle surprised you most? Was there one that felt obvious when you read it but turned out to be harder to follow in practice? That's usually the one worth paying the most attention to.
Your charter will evolve as your relationship with AI matures — and that's exactly the point. The goal isn't to get it perfect today. It's to have something written down that makes tomorrow's decisions a little clearer. You've already done the hardest part: thinking carefully about what kind of AI user you want to be.