Detailed content
🧭 New here? 4 words before you start
- AI / LLM: an "LLM" (large language model) is the type of AI that runs behind ChatGPT, Claude, or Gemini. You chat in text, and it responds in text.
- Prompt: is the message you write to AI. “Good prompt” = a clear request with context, not a question out of the blue.
- Context: the information you provide along with your request (your data, examples, rules). It’s what makes the answer tailored to you.
- Hallucination: when AI makes up a data point that looks true. It happens—that’s why human review is part of the job.
🌍 AI isn't just for tech professionals
There’s a myth that holds a lot of people back: that “AI is for programmers.” It’s false. You you don’t need to write a single line of code to get a real advantage from AI. People who apply AI to concrete tasks in their own fields—writing, sales, planning, customer service, legal, HR—get ahead of those who wait for “someone in IT” to solve it.
What is
Treat AI as a general-purpose tool that lives within your work, not as a separate technical subject. The advantage comes from applying AI to area you already know — where you know what a good answer looks like and what’s nonsense.
✓ What to DO
- ✓Take a task you already know how to do and use AI to do it faster.
- ✓Start with your own work, where you can judge the result.
- ✓Measure time: did what used to take 1h take 20 min?
✗ What NOT to do
- ✗Wait until you’re “technical” to start using it.
- ✗Use AI in a field you don't understand — you won't know whether the answer is any good.
- ✗Outsource to “the IT folks” what is your responsibility.
Why learn
Because the most useful AI is the kind that fits into your current workflow, not the kind that requires you to become someone else. The professional who knows how to use AI in their own field increases their own value without changing careers. And there's evidence for it: in customer support, AI increased average productivity by +14% — and by +34% for beginners, because it spreads the best practices of the most experienced.
📊 Why this matters (data)
+14% on average and +34% for beginners productivity in customer support — a peer-reviewed study of 3 million real conversations (Brynjolfsson, Li & Raymond, NBER w31161, published in Quarterly Journal of Economics). The benefit is real and greater for those just starting out.
Key concepts
Domain > code
Real task > demo
+34% for beginners
Initiative > waiting
⭐ Become the team's “AI person”
Almost every team has an open position: the person who solves real problems with AI before others do. It’s not the biggest technical expert — it’s the person who brings a ready-to-use solution to the meeting. You earn that position through small, visible wins, not a degree.
What is
Become the team’s go-to practitioner for applying AI to concrete, everyday problems. The key is solve and show: you take a recurring pain point for the team (a tedious report, a meeting summary, a standard response) and deliver a working solution.
How it works in practice (4 steps)
Choose a small, recurring pain point
Something the team does every week that nobody likes. The more repetitive it is, the greater the perceived gain.
Solve it yourself, with AI, on the side
Don’t promise anything. Iterate until it truly works on the real task (not in a demo).
Show the result, not the process
“Look, I automated the meeting minutes summary—it saves 2h a week.” Results speak louder than jargon.
Teach the next person
Sharing what worked multiplies its value and establishes you as the team’s go-to person.
Why learn
Because in a market where 63% of companies point to the lack of qualifications as the biggest barrier (WEF, 2025), being the person who helps the team move forward with AI is scarce career capital. It’s not about knowing everything — it’s about solving problems with AI before of others and make it visible.
📊 Why this matters (data)
63% of employers cite the skills gap as the biggest barrier to transformation by 2030, and 85% plan to prioritize upskilling (WEF Future of Jobs Report 2025). Whoever takes on the “AI person” role fills exactly this gap.
Key concepts
Solve > explain
Small wins
Make visible
Teach the next person
⚖️ Use AI with human judgment
AI generates answers convincing — not always correct. Accepting any output without reviewing it is the fastest way to be confidently wrong. The habit that separates the amateur from the professional is simple: review, adapt it to your style, and decide whether the result deserves to bear your name.
💡 What is “ground truth”?
“Ground truth” is the actual answer key — the truth you use to check AI’s answer. When AI says “this is 92% correct,” the first question to ask is: “is the answer key it used really true?” Trusting the sales material’s metric instead of questioning the ground truth is the #1 evaluation failure (MIT Sloan Management Review).
What is
Keep humans in control of the decision. AI proposes; you review, correct, adapt, and owns the result. You are the editor responsible, not a rubber stamp. This applies especially to anything involving data, numbers, proper names, or factual claims.
Practical example (copy and use)
Objective: ask the AI itself to list where it might be wrong before you trust the answer.
How to verify: the answer should include a list of risky claims and what to check; if it says "everything looks good, trust it," the prompt failed.
Aja como um revisor cético do texto abaixo.
Não reescreva. Em vez disso:
1. Liste cada afirmação factual (número, data, nome, citação).
2. Para cada uma, diga: "verificável?" e "como eu confirmaria?".
3. Marque o que pode ser alucinação ou suposição minha.
4. Termine com 1 frase: o risco mais alto se eu publicar isto sem checar.
TEXTO:
<cole aqui o texto que a IA gerou>✓ Human judgment
- ✓Check numbers, dates, and names against the source.
- ✓Adapt the tone to your context and audience.
- ✓Only publish what you’d stand behind in a meeting.
✗ Automatic stamp of approval
- ✗Copying and pasting the answer as if it were true.
- ✗Believe the metric the tool itself promotes.
- ✗Blaming "AI" for the error — your name is on the deliverable.
Why learn
Because for critical work 90% of people still prefer humans (MIT NANDA)—exactly because someone needs to make sure it’s right. Judgment is what AI doesn’t provide, and what keeps you in control.
Key concepts
Convincing ≠ correct
Check the ground truth
You are the editor
You sign
🧩 Context engineering
This is the biggest lever of all. The difference between a mediocre answer and a tailored answer almost never comes down to “knowing the magic prompt” — it comes down to the context you provide. Instead of having a generic conversation, you give AI real information: your projects, priorities, examples of how you work, and the rules that matter.
💡 What is "context engineering"?
It's the habit of deliberately assemble the information the AI receives along with the request — role, real data, examples, constraints, and output format—instead of sending a question out of the blue. “Engineering” here doesn’t mean programming: it means organizing what goes in. AI is only as good as the context you give it.
How to read: on the left, a loose request goes in and a generic response comes out. On the right, the four context blocks (cyan)—role, real data, an example of your style, and the expected format—converge into a tailored response (green). Context is what changes the outcome, not the “magic word”.
Practical example (copy, replace the fields, and run)
Objective: turn a vague task ("help me with X") into a request with context that produces a usable response on the first try.
How to verify: the answer should sound like you (not generic), follow the format, and use the data you pasted. If it comes out neutral and "by the book," it lacked context.
# PAPEL
Você é <seu cargo, ex.: analista de marketing> e escreve no meu lugar.
# OBJETIVO
<o que eu quero, em 1 frase, ex.: um e-mail pedindo prazo extra a um cliente>
# CONTEXTO REAL (use isto, não invente)
- Público: <quem vai ler>
- Situação: <o que está acontecendo>
- Dados/fatos: <cole números, nomes, datas reais>
# COMO EU FALO (imite este tom)
<cole 1 exemplo curto de algo que você já escreveu>
# RESTRIÇÕES
- Tamanho: <ex.: até 120 palavras>
- Não usar: <jargão, promessas que não posso cumprir>
# SAÍDA
Entregue só o texto final. Depois, em 1 linha, diga o que você assumiu.💡 Practical tip
Save this framework in a notes app and reuse it. In a few weeks, you’ll have a “base context” for your work that helps any AI respond in your style—it’s the asset that saves you the most time day to day.
Why learn
Because the root cause the MIT identified for so many AI projects failing is “learning gap”: the tool forgets the context and requires you to provide it every time. People who master context engineering work around exactly this bottleneck—and get much better results from the same AI than their neighbor does.
Key concepts
Real data > generic prompt
Example > vague instruction
Role + format
Reusable baseline context
Self-recovery (optional): which of these is the best use of context engineering?
🔁 Work iteratively
Stop trying to get everything right on the first try. With AI, speed comes from short cycle: make a quick draft, test it, see what’s wrong, fix it, and repeat. Each round teaches you what to ask for next time. It’s the opposite of getting stuck searching for the perfect prompt.
What is
Treat work as a loop instead of a straight line: draft, test, evaluate, and correct, repeating the cycle. Each pass builds on what you learned in the previous one — you don’t start from scratch; you improve the version you already have.
How to read: the work isn’t a straight line to perfection—it’s a loop. The cyan arrow pointing back is the secret: each loop carries what you learned and makes the next draft better. A short cycle beats a perfect first try.
✓ Iterate
- ✓Deliver a “good enough” draft and improve it with feedback.
- ✓Tell AI what didn’t work and ask it to fix only that.
- ✓Treat mistakes as information, not failure.
✗ Paralysis
- ✗Rewriting the prompt 10 times before testing it once.
- ✗Wait for the perfect version before showing it to anyone.
- ✗Starting over from scratch with every adjustment instead of correcting.
Why learn
Because AI makes short cycles cheap — generating the next version takes seconds. Those who work by iterating take advantage of this; those who seek initial perfection throw that advantage away. A good result emerges from the cycles, not on the first try.
Key concepts
Draft > perfection
Short cycle
Error = information
Feedback leads to corrections
⚙️ Automate predictable tasks
Look at your week and mark what is repetitive and predictable: the same report, the same standard response, the same spreadsheet. That’s a candidate for automation. But there’s one part that almost no one teaches: knowing when NOT to use AI. If a task follows a fixed rule, simple automation is cheaper, more reliable, and easier to audit than AI.
What is
Two moves together: identify what is repetitive and predictable to take off your plate, and choose the right tool for each case—sometimes a simple rule, sometimes AI. Automating well is as much about knowing how to use AI as knowing when it’s overkill.
💡 Fixed rule vs. AI
One “simple automation” is a sequence of fixed steps (e.g., “every Friday, get the spreadsheet, add up column C, and send it by email”). It doesn’t think, doesn’t make creative mistakes, and costs almost nothing. Use AI when the task has ambiguity — free-form text, judgment, cases that vary. Deterministic task → rule. Ambiguous task → AI.
How to read: before "throwing AI at the problem," ask whether the task always follows the same rule. If yes, simple automation is cheaper. If the task is ambiguous, that’s where AI shines — always with human review.
📊 Why this matters (data)
Gartner predicts that 40%+ of agentic AI projects will be canceled by 2027 — because of rising costs and uncertain value (it’s a prediction, not an established fact). In practice: throwing AI at everything, including tasks a simple rule could handle, is expensive and risky. Choosing the right tool for each task is part of the job.
Why learn
Because well-chosen automation gives you hours back in your week for the work that matters — and knowing when not to use AI avoids cost, risk, and fragility. Maturity with AI means knowing when to use it and when it’s not worth it.
Key concepts
Predictable → rule
Ambiguous → AI
Lower cost, lower risk
Saying no is also a skill
🛡️ Build professional and financial security
The previous six habits make you more productive. This one turns productivity into safety. The idea: use AI to build new sources of income around a field you already know well, reducing your dependence on a single job. Productivity that stays only with the company disappears if the job does; productivity that becomes your own asset does not.
What is
Use AI's leverage to create optionality: a side business, a digital product, a small consulting practice, content that generates income. It’s not about "quitting everything"—it’s about building, little by little, something that depends on you and not just on a single paycheck.
✓ Build confidence
- ✓Package what you know into a small service or product.
- ✓Use AI to do on your own what used to require a team.
- ✓Start small, alongside your job, and iterate.
✗ Pitfalls
- ✗Fall for “magical passive income with AI” (classic grift).
- ✗Quitting your job before validating any income.
- ✗Building something in a subject you don’t truly know.
Why learn
Because the market is changing: the WEF projects +78 million net jobs by 2030, but with 22% of current jobs facing disruption, with ~39% of skills changing. In this landscape, having more than one leg to stand on isn’t a luxury—it’s prudence. AI has lowered the cost of building that second leg.
📊 Why this matters (data)
+78M net jobs by 2030 (170M created, 92M displaced), 22% of jobs in disruption and ~39% of skills changing (WEF Future of Jobs 2025). Change creates opportunity for those who position themselves—and risk for those who depend on a single source.
Key concepts
Multiple income streams
Proprietary asset
Optionality
Start small
🏁 Module summary
Next step:
Track 2 · Be irreplaceable — the human skills AI can’t replace. Using AI to amplify them is halfway there; together, they make up the complete skill set.