You’re Not Bad at AI. You Need Better Questions.
Most people are not bad at AI. They are handing the tool vague requests and getting vague answers back.
You know the prompts.
“Write this better.”
“Give me ideas.”
“Help me with my business.”
“Make this sound good.”
Technically, AI can answer all of them. The problem is that it has to guess almost everything.
Who should it act as?
Who is the answer for?
What are you actually trying to create, solve, or decide?
What should it include?
What should it leave out?
How should the answer be shaped?
Then you get the answer back and think, “Fine. I’ll just rewrite it myself.”
That is where most AI frustration starts.
Not because the tool is useless. Not because some 24-year-old on YouTube unlocked a secret level the rest of us can’t reach because he owns a ring light and says “leverage” too much.
The reason is usually simpler.
You are starting from a blank prompt every time, and the blank prompt makes the tool do too much guessing.
At a Glance
- Most people are not bad at AI. They are handing it vague requests and getting vague answers back.
- The fix is not a secret prompt or another subscription. It is context.
- Five plain questions turn a vague request into one worth reviewing: who AI should act as, what you want done, why you need it, what context it needs, and what you need to check before you use it.
- Better questions get stronger first drafts. They do not replace your judgment, and they do not make AI right.
- Start with one real request and the free guide, Say It Right the First Time.
1. AI is not a mind reader
AI can be useful. Useful does not mean psychic.
Give it a vague request, and it fills in the gaps with its best guess. Sometimes the guess is harmless. Sometimes it sends the answer somewhere you never wanted it to go.
You asked for a blog intro and got a sales-page monologue. You asked for an email and got something that sounds like a software company apologizing for a data breach. You asked for “plain English” and got a sentence wearing a blazer.
Then people say, “AI does not work for me.”
Maybe. Or maybe it did exactly what it was told, and it just was not told enough.
That is not a character flaw. It is a structure problem. Structure problems are fixable.
2. The vending-machine prompt
A lot of people use AI like a vending machine. Put in a vague request. Push the button. Hope something useful drops. Then get annoyed when the bag of chips gets stuck behind the glass.
There is nothing wrong with a short prompt. Short is fine. Underbuilt is the problem.
A weak prompt does not tell the tool enough about the job. It makes AI guess the role, the audience, the purpose, the context, the format, and the standard all at once. That is a lot of guessing for a tool people already accuse of making things up.
A better prompt does not have to be long just to look impressive. We are not writing a hostage note to a robot. It just has to be clear.
You are not trying to make AI magic. You are trying to make it useful. Much less glamorous. Much more valuable.
3. You need context before you need more tools
The AI world keeps trying to sell you another tool. Newer. Smarter. Faster. Now with a dashboard and a name that sounds like a medication commercial.
Some tools are useful. But if you cannot give AI a clear job, a new tool just hands you a shinier version of the same generic answer. Now it is not just beige. It is beige with animation.
The first upgrade usually is not another subscription. It is context. A way to slow down the request before you send it.
That is the whole idea here. Not magic. Not a secret prompt. Not a prompt library so large you need a search bar and emotional support. Just a repeatable habit you can use on a Tuesday morning without joining a Discord or pretending “prompt engineering” needs to sound like a federal agency.
4. The five things AI needs from you
Before you ask AI for the answer, give it five things. You do not need to label them or fill out a form. You just answer them somewhere in your request.
Who should AI act as for this task? A generic assistant answers like a generic assistant. Point it at the role that actually does this work.
What do you want done? Say the actual task plainly, not “help with this.”
Why do you need it? The reason behind the request changes the answer you get.
What context does AI need? This is the one most people skip and the one that matters most. The facts, limits, audience, tone, and details it has no way of knowing on its own.
What do you need to review, verify, or decide? That part stays with you. The tool produced a draft. You decide what is true and what is worth keeping.
That is it. One repeatable way to stop staring at the blank box and hoping the tool figures you out. It will not. It has many talents. Reading your mind is not one of them.
5. What a weak prompt looks like
Here is the kind of prompt many people start with.
Weak prompt: Write me a checklist for bringing my mom home from the hospital.
That is not a terrible request. It is just thin. The tool does not know who the checklist is for, what tone you need, what to include, what to avoid, how detailed it should be, or where the medical boundary is. So it guesses. Sometimes the guess is useful. Sometimes it is not.
Now compare that with a request that gives the tool some context.
6. What a better prompt looks like
Stronger prompt:
Act as a plain-English caregiver checklist writer helping a family organize hospital discharge questions. Do not give medical advice, diagnosis, or treatment instructions.
I want to turn my rough notes into a simple one-page checklist called “Before Mom Comes Home From the Hospital.” It is for adult children preparing to bring an older parent home. They are overwhelmed and afraid they will forget something important. They are not medical professionals, so the checklist should help them prepare better questions before discharge.
Include medication changes, follow-up appointments, warning signs to ask about, home safety, equipment needs, and who to call after hours. Do not diagnose, give treatment advice, or make promises about outcomes. Keep the tone calm, practical, and plain English.
Give me a short introduction under 100 words, a simple checklist with clear headings, five questions to ask before leaving the hospital, and a short reminder to contact the care team for personal medical concerns.
That is a different assignment. Not because it is fancy. Because it tells the tool what job it has.
The weak prompt asks AI to guess. The second one gives AI a lane. That is the difference.
7. Context does not make AI perfect
A better request improves clarity. It does not make the answer true. It does not replace your judgment. It does not replace fact-checking. It does not replace advice from someone actually qualified for your situation.
AI can sound confident while being incomplete, outdated, or wrong. Sometimes it is wrong with excellent formatting. That is the dangerous kind. Very persuasive. Very wrong.
So before you use what it gives you, read it like the operator, not the audience. Ask:
- Did it answer the actual request, or did it wander off?
- Did it assume something I never gave it?
- Did it ignore a constraint that mattered?
- Are the facts right? Does anything need a source?
- Is this a legal, medical, financial, or safety decision that needs a qualified human?
- Did it make the answer smoother but less useful?
Use AI as a tool. When money, health, business, or people you love are involved, the judgment stays with you.
8. Why this matters for regular people, not “AI people”
The internet has made AI feel more complicated than it needs to be.
People sell giant prompt libraries. People act like every new tool will change your life by lunchtime. People post screenshots of their “AI stack” like it is a family portrait.
That is the circus. Most Gen X adults do not need theater.
They need AI to help with normal work. Drafting clearer emails. Summarizing messy notes. Comparing options. Preparing questions. Stress-testing a decision. Turning rough thoughts into something usable.
You do not need to become an AI person. You need a repeatable way to give the tool a real assignment, and then the judgment to check what it hands back. That is the whole game.
The LCR Bottom Line
You are not bad at AI. You are probably asking the tool to do too much guessing. That is fixable.
Not with a secret prompt. Not with another guru thread. Not with a 400-prompt library you will forget you downloaded.
With context. Tell AI who to act as, what you want, why you need it, and what it is working with. Then read the answer with a clear eye and keep what is worth keeping.
That is how you give AI a real assignment. That is how you get a stronger first draft without pretending the tool is the authority.
AI can help. Your judgment still leads.
Want the whole habit on one page you can actually keep?
Say It Right the First Time: A No-Nonsense AI Guide for Getting Better Answers is a free Lisa C Recommends guide. It walks through the five questions in plain English, with real examples and a one-page template you can paste straight into whatever AI you use. No jargon, no hype, no prompt-bro nonsense.
Start with one real request, the kind you were already going to make, and see whether a little context changes what you get back.