Most people who "use AI" are stuck at Level 2 — the Curious stage. They've tried ChatGPT, maybe a few other tools, and they've seen what AI can do. But they don't have a consistent habit. They don't have prompts that reliably work. And they're not sure what to actually learn next.
The jump to Level 3 — Tinkerer — is the most important skill step you'll make. It's the difference between AI being a novelty you reach for occasionally and a tool you actually get value from every week.
Here's exactly how to make that jump in 30 days.
What separates Curious from Tinkerer
The Curious persona (Level 2) uses AI reactively. Something reminds them AI might help, they try it, they get a result that's okay, and they move on. There's no system. No iteration. No accumulated learning.
The Tinkerer persona (Level 3) uses AI proactively. They have at least 3–5 tasks they run through AI every week, they've iterated on those prompts until they work reliably, and they've started building a personal playbook of what works for them.
The difference isn't intelligence or technical skill. It's consistency and deliberate iteration.
The 30-day playbook
Week 1: Find your anchor task
Pick one task you do every single week. Not "something you could use AI for" — something you already do that AI could help with.
Good candidates:
- Weekly summary email to your manager
- First draft of a recurring report
- Social media captions for the week
- Briefing a colleague on a project update
- Researching a topic before a meeting
This is your anchor task for the next 30 days. Everything else is optional experimentation. This is the one thing you commit to doing with AI every time for a month.
Why one task? Because consistency is how habits form, and habits are how skills develop. Trying to use AI for everything immediately means you never get good at using it for anything.
Week 2: Build your first real prompt
Most people write prompts like this: "Write an email update about the project."
That produces generic output. It produces output that sounds like everyone else's AI-generated emails. And it produces output that requires heavy editing before it's usable.
A good prompt has three parts:
- Role and context — who is producing this output, and for whom
- The actual task — specific, with constraints
- Format and tone guidance — what it should look like when it's done
Here's the same prompt rewritten:
I'm a product manager writing a weekly update for my CEO. The tone is clear, direct, and under 200 words — no jargon, no filler. This week's update covers: [3–4 bullet points]. Write the update.
Night and day difference.
Spend Week 2 iterating on your anchor task prompt until you'd send the output to a real person without major edits 8 times out of 10.
Week 3: Add a second task and start a prompt library
Once your anchor task prompt is solid, add a second task. Apply the same three-part structure.
Start keeping a simple document — Notes app, Notion, anywhere — called "Prompts that work". For each prompt, note:
- What it's for
- The exact prompt
- Any variables you change each time (highlighted or in brackets)
This is the beginning of your personal playbook. By the end of 30 days you'll have 2–4 solid prompts you actually use. That's enough to put you in the top 20% of people who "use AI".
Week 4: Iterate and stress-test
Take each of your working prompts and deliberately try to break them.
Ask: what happens when I change the context? What if I need a different tone? What if the input is messier than usual?
Push each prompt until you find its edges, then refine it to handle the variations you'll actually encounter. This is what separates a prompt that works once from a prompt that works reliably.
By the end of Week 4, you've done the work. You have a real habit, a real playbook, and real evidence that AI is genuinely useful for you.
What Tinkerers know that Curious users don't
A few things you'll have learned by the end of this month:
Specificity beats creativity. The more specific your prompt, the better the output. "Write a marketing email for a B2B SaaS product for HR managers who struggle with employee retention, aiming for a 35–45% open rate, with a focus on pain over features" beats "Write a marketing email" every time.
Context is free. You can give AI as much background as you want. Don't ask it to guess — tell it. Your company, your audience, your tone of voice, the specific outcome you need. All of it.
The first output is a draft, not a result. Tinkerers don't accept the first response. They iterate. They ask AI to make it shorter, make it sharper, change the angle, try a different structure. The best output is usually 2–3 rounds in.
Your prompt library is an asset. Every solid prompt you build is time you save forever. A library of 20 working prompts is worth hours per week.
After the 30 days
If you've followed this playbook, you're a Tinkerer. You have:
- A consistent daily/weekly AI habit
- 3–5 prompts that reliably work
- A personal playbook you're adding to
- Demonstrably better results than "ad-hoc AI user"
The next level — Craftsperson — is about applying this same rigour to systems rather than individual tasks. Instead of one good prompt, you build a whole workflow. Instead of helping yourself, you start building things that help your team.
But that's the next step. For now: one anchor task, one working prompt, 30 days.
Take the quiz to find your current level →
Or read about what the Tinkerer persona looks like in full — the Scientist is the most analytically-minded Tinkerer, but all four L3 personas share the same "building a real habit" foundation.