Most People Are Using AI Wrong — Here’s What Actually Works
AI isn’t underdelivering. Most people are just approaching it the wrong way. Here’s how to fix that.
I’ve watched it happen dozens of times. Someone tries an AI tool for the first time, types something like “help me with my essay” or “write me some code,” gets a mediocre result, and walks away convinced that AI is overhyped. They tell their friends it’s not that impressive. They go back to doing things the old way.
And honestly? I get it. Because when I first started using AI tools, I did the exact same thing.
The problem isn’t the technology. The problem is how most people approach it — and once you understand the difference, the results you get become almost unfairly good.
The Mistake Everyone Makes First: Being Too Vague
This is the big one. The single habit that separates people who get extraordinary results from AI and people who get generic slop.
Most people treat AI like a vending machine. You put in a vague request, you expect a perfect result to come out, and when it doesn’t you blame the machine. But AI tools — especially the best ones like Claude — are much more like a brilliant collaborator than a vending machine. And you wouldn’t walk up to a brilliant collaborator and say “help me with my thing” and expect magic.
Compare these two prompts:
Vague: “Write me a cover letter.”
Specific: “Write me a cover letter for a software engineering internship at a fintech startup. I’m a second-year computer science student with experience in Python and React. I want to sound confident but not arrogant, and I want to mention that I built a personal budgeting app as a side project. Keep it under 300 words.”
The second prompt takes 30 extra seconds to write. The output is night-and-day different. That’s not the AI getting better — that’s you getting better at using it.
The rule of thumb: give the AI the same context you’d give a smart human who knew nothing about your situation. Role, goal, constraints, tone, length, audience. The more specific you are, the less it has to guess — and AI guesses toward the average when it doesn’t have enough information. Average is forgettable. Specific is powerful.
Mistake #2: Accepting the First Answer
AI tools are not oracles. The first response is a starting point, not a final product.
The people who get the best results treat AI like a draft machine, not a delivery service. They take the first output, identify what’s missing or off, and push back. “Make this more concise.” “The third paragraph is too generic — can you make it more specific to X?” “I don’t like the tone — can you make it sound less formal?”
This back-and-forth is where the magic actually happens. A great AI interaction isn’t one prompt and done — it’s a conversation. Think of it like working with an editor. You wouldn’t hand a first draft to an editor and expect them to hand back a perfect final piece. You’d go back and forth. Same principle applies here.
Claude in particular is very good at this. It handles follow-up instructions well, maintains context across a conversation, and will tell you honestly when a direction isn’t working. Use that. Push it.
Mistake #3: Using AI as a Search Engine
Type a question in, read the answer, close the tab. That’s a search engine workflow, and it massively underutilizes what AI is actually capable of.
The difference between a search engine and an AI assistant is interactivity. A search engine returns links. An AI assistant can reason with you, adjust to your specific context, challenge your assumptions, and help you think through something you haven’t fully figured out yet.
Some of the most valuable ways to use AI have nothing to do with getting an answer — they’re about working through a problem. “I’m trying to decide between these two approaches — what are the trade-offs?” “Here’s my argument — what’s the strongest counterargument?” “I think this code is inefficient but I’m not sure why — walk me through it.”
That kind of usage turns AI into a genuine thinking partner. And a thinking partner is worth a hundred times more than a fancier search engine.
Mistake #4: Using It to Replace Thinking Instead of Enhancing It
This one is a little more uncomfortable to say, but it needs to be said.
A lot of people use AI as a shortcut to avoid thinking. Copy the essay, submit the code, take the output and run. And look — no judgment, the temptation is real. But this is also where AI use goes from powerful to actively harmful to your own development.
When you use AI to skip the thinking, you don’t build the skill. You don’t develop the judgment to know when the output is wrong, which it sometimes is. You don’t improve. And ironically, you end up more dependent on the tool, not less — because you never built the foundation underneath it.
The better model — especially for students — is to use AI to augment your thinking. Draft your own argument first, then ask Claude to poke holes in it. Write your own code first, then ask it to review it. Form your own opinion, then use AI to pressure-test it. You stay in the driver’s seat. The AI makes you sharper, not lazier.
Mistake #5: Not Giving It a Role
One underrated trick: tell the AI who to be.
“You are a senior software engineer reviewing my code for a production environment.” “You are a skeptical editor who is hard to impress.” “You are a professor explaining this concept to a first-year student.”
This sounds almost too simple, but it meaningfully shifts the quality and angle of the response. AI models respond to framing. Giving Claude a role gives it a lens to interpret your request through — and that lens changes everything from vocabulary to depth to tone.
Try it once and you’ll never go back to roleless prompting.
Mistake #6: Giving Up After One Bad Experience
AI tools have improved at a pace that’s genuinely hard to keep up with. If you tried one a year ago and walked away unimpressed, the version you tried and the version available today are not the same thing.
Claude in particular has seen significant improvements in reasoning, coding, and long-form writing. The tools are better. But more importantly — if you’ve internalized even half of what’s in this article, you’re better at using them than you were. And that matters more than any model update.
The Simple Framework That Changes Everything
If you want to remember one thing from this article, make it this:
Be specific. Push back. Have a conversation.
That’s it. That’s the gap between people who find AI transformative and people who find it disappointing. It’s not about which tool you use or which tier you pay for. It’s about showing up with context, staying engaged, and treating the interaction like a collaboration rather than a transaction.
The technology is extraordinary. Most people just haven’t figured out how to meet it where it is.
Now you have.
Follow for more on AI tools, student productivity, and actually useful takes on where technology is headed.
Most People Are Using AI Wrong — Here’s What Actually Works was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.
