AI Curious

The hardest step is the first one. These get a tool open and into the work you already do.

• Create a free account on one AI assistant — ChatGPT, Claude, or Microsoft Copilot — and have one real back-and-forth (not just a single question). If work only allows its own AI tool, use that one at work — and a free private account for everything else. What separates people who "get" AI from people who don't usually isn't talent — it's that one group actually opened the tool and kept talking. Everything else builds on this.

  Getting Started with AI ↗ one-time setup
• Use AI for one real task from your actual week — not a test, something you genuinely need done. It clicks when it solves a real problem. A demo you'll forget; your own Tuesday-afternoon headache you'll remember.

  Try this: Paste a long email thread and ask it to summarize and flag what actually needs a reply. Or take a topic you've never understood and ask it to explain from scratch.

  Prompt Summarize this email thread and tell me what I actually need to reply to. Copy

  Getting Started → find your use cases ↗ keep doing it
• Learn the three moves you'll use constantly: start a new chat, ask a follow-up in the same chat, and paste or upload a document for the AI to work from. Most early frustration is mechanical, not conceptual — people open a fresh chat every time and throw away all the context they just built.

  Try this: The one that trips people up: stay in the same chat while you're on one topic so the AI keeps the thread. Start a new chat only for something genuinely different.

  keep doing it

Master this and most "AI is useless" frustration disappears. It nearly always comes down to one thing: context.

• Picture AI as a brilliant new hire: it shows up knowing almost everything in general, and almost nothing about you, your job, or your company — until you tell it. This one picture explains nearly every bad answer you'll get. The AI isn't dumb; it's a sharp stranger missing the brief that only you can give it.

  ✓ Tick it when you can say, in a sentence, why AI hands you generic answers — and what to add to fix it.

  Try this: See it for yourself: ask "what should I focus on this week?" with no other context. The vague answer you get back is the new hire guessing, because you never briefed them.

  AI Knows Nothing About Your Company ↗ one-time setup
• Give the AI context in your request: who it's for, what you're trying to achieve, and the format you want back. The gap between a useless answer and one you can actually use is almost always context, not clever wording. Highest-return habit at this stage.

  Try this: Watch the difference. Weak: "write a product description." Strong: the briefed version below — same task, far better result.

  Prompt Write a 50-word product description for busy hospital nurses — friendly, not salesy — for our shift-scheduling app. Copy

  AI Knows Nothing About Your Company ↗ keep doing it
• Try reverse prompting once: before it answers, tell the AI to ask you a few questions that would help it do a better job. You don't need to write the perfect prompt — you can make the AI pull the details out of you. The single biggest unlock for anyone who "doesn't know what to ask."

  Try this: Paste this line before your actual request, then answer the questions it asks back.

  Prompt Before you answer, ask me up to 5 questions that would help you give a better response. Copy

  Let AI Interview You ↗ keep doing it
• Iterate instead of starting over: when an answer is off, tell the AI exactly what's wrong and let it revise. The first reply is a draft, not a verdict. People who treat AI as a conversation get far better results than people who treat it as a vending machine.

  Try this: Don't restart — steer. "Shorter, and cut the intro." "Too formal — write like I'd message a colleague." "You missed the budget constraint."

  Working With AI, Not Through It ↗ keep doing it

This is where the mental model becomes a setting. Instead of giving context in every chat, you tell the tool who you are once and it remembers. ChatGPT, Claude, and Microsoft Copilot all have this — the menus differ, the idea doesn't. A few bullet points is plenty here.

• Turn on memory in your main AI tool so it remembers things across chats. With memory off, every conversation starts from zero and you re-explain yourself forever. Flip it on once and the tool starts compounding what it learns about you.

  Try this: It's usually under Settings → Personalization, though the exact label varies by tool. Turn it on, tell it one true thing about you, and check it remembers in your next chat.

  The Setup That Makes AI Stop Being Generic ↗ one-time setup
• Add a few bullet points about yourself to your custom instructions — your role, where you're based, the languages you work in. This is the new hire's standing brief. A few lines here and every future answer already lands in your language, your context, your units — without you asking.

  Try this: Drop a few lines like this into your custom instructions:

  Prompt I'm a marketing lead in Zurich. I work in German and English. Keep answers concise and skip the corporate fluff. Copy

  The Setup That Makes AI Stop Being Generic → Tier 1 ↗ one-time setup
• Set one style preference you keep fighting for — "be concise", "skip the corporate filler", or "push back when I'm wrong". Most people re-fight the same battle in every chat: too long, too fluffy, too agreeable. Set it once in the preferences and stop.

  Try this: Add the instruction you re-type most to your custom instructions. Something like:

  Prompt Be concise. Skip the corporate filler. Push back when you think I'm wrong instead of just agreeing. Copy

  Your AI Agrees With You Too Much ↗ one-time setup

Two habits that stop AI from quietly costing you. Build them now, while the stakes are low.

• Catch a hallucination in the wild: ask AI about something you already know well, and spot where it states something false in a confident tone. Once you've seen AI be fluently, confidently wrong about something you know, you'll never blindly trust it again. That instinct is worth more than any prompt trick.

  Try this: Ask about your own field, your hometown, or a niche hobby — somewhere you'll instantly catch a confident wrong answer.

  Getting Started → always verify ↗ keep doing it
• Set your own data rule: don't paste confidential, client, or personal information into free AI tools — and remember the "about me" profile you just set up lives in that same account. Free tiers often use what you type to train future models, and the profile you set up sits in that account too. One clear rule now saves a bad surprise later.

  ✓ Tick it when you know what you'd happily paste and what you'd keep out.

  Try this: Rule of thumb: if you'd hesitate to post it in a company-wide channel, don't paste it into a free AI tool.

  Getting Started → don't share sensitive data ↗ keep doing it

Skills fade without reps. This is how the foundation actually sticks.

• Use AI for something — anything — every day for one week. Fluency forms through reps, not study. Week one feels clunky; by week three the results get genuinely useful and the habit starts to hold.

  Try this: Keep it stupid-small: one email, one summary, one question a day. Momentum beats ambition.

  How to Actually Get Good at AI ↗ daily habit
• Pick one source to keep learning from, so you're not relying on random social-media noise. (This site's newsletter counts.) The field moves fast and the temptation is to follow 50 influencers and feel buried. One trusted source beats the firehose every time.

  Join the newsletter ↗ now and then