Open the syllabus of almost any AI course and the first thing you find is the tool. Here’s the interface, here are the functions, here are the prompts. It’s the natural order, the one any software has always been taught in: this is the instrument, here’s how you use it. Three days later you know how to make the tool do things. Then you sit down in front of a real problem — that quote, that negotiation, that decision that doesn’t add up — and you’re stuck. The course answered a question you hadn’t asked yet.
The tool is an answer
Because with AI that order, which works elsewhere, jams here. The tool is an answer: it’s the thing you pick up after you’ve figured out what you need. With traditional software the two moments coincide, because the tool does one thing — the spreadsheet calculates, the management system manages. AI doesn’t: it does almost everything and nothing in particular, and that “almost everything” becomes something useful only when it arrives with a precise problem to solve. Without the problem in front of you, you learn to operate a tool that can go in a thousand directions, without knowing which one is yours.
Where it stalls
You see it at the moment of the handoff. With the tool open, in a guided session, everything flows: you click where you’re meant to click, you write the prompt you watched someone write. Then you go back to your desk, in front of something that matters, and you stop. Not because you can’t use the tool — you use it well. Because the part that decides comes first, and it was left out: what am I trying to achieve, for whom, what’s the real obstacle. That stuff isn’t inside the tool, and a path built on the tool never runs into it. Not out of bad faith on the teacher’s part: it’s simply not what that kind of course set out to teach.
And the tool doesn’t correct confusion: it executes it. Give it a hazy goal and it carries it out to the letter, fast, confidently, ready to send. The more capable it is, the more treacherous — because it executes a crooked request flawlessly, and the result looks fine until it’s too late. It’s easy to believe the challenge is finding the right command. The challenge was one step earlier: deciding what to ask. And no command saves a question that was never asked.
The order is what lasts
There’s also a concrete reason the order weighs more than the tool. The tool changes — new versions, new capabilities. The sequence doesn’t. “Problem first, then maybe the tool” holds for this tool and the next one, for this version and next year’s. It’s the part that stays intact when the rest updates. Learn a tool, and you know that one. Learn to start from the problem, and you can use them all — even the ones that don’t exist yet.
System first, then the app
A course that puts the order right does the opposite, and it’s neither better nor nobler: it answers a different question, the one that counts when you go back to work. It starts from your problem — what you need to achieve, for whom, where you get stuck right now — and the tool comes after, chosen for that problem, sometimes discarded because for that problem it wasn’t needed. It’s the difference between strategy and execution: the tool is an application, the strategy is the system it runs on. An application with no system underneath runs anyway, but it does the wrong thing efficiently. System first, then the app.
From here, a useful question before choosing a course — not to find out who’s good and who isn’t, but to understand what you’re buying. Where does it start: from the tool, or from my problem? You can read it in the title and the table of contents. If they start from the tool — “AI for [something],” “master [the tool]” — the course will teach you to use it, and it can do that very well: it’s useful if what you’re missing is practice with that tool. If they start from a result or a decision, you’re getting closer to the competence you need when the tool, on its own, isn’t enough. They’re two different things, both legitimate — the mistake is buying the first while believing you’re taking home the second. The same test works for you every day, and it costs nothing: before opening the tool, try writing the problem in one sentence. If it won’t come, it’s not the tool you’re missing.
The tool is the last decision, not the first. You choose it when you know what you’re trying to do, not to find out. Because the most powerful tool in the world, aimed at a question you didn’t ask, works perfectly: it gives you the right answer to the wrong problem.