You've tried AI. You've typed a question, gotten a wall of text back, and thought: this isn't that useful. Maybe it was technically correct but completely generic — the kind of answer that could apply to any business in any industry anywhere. You're not wrong to be frustrated. But the problem probably isn't the AI.
It's the prompt.
Here's the thing nobody tells you when you sign up for Claude or ChatGPT: these tools are designed to be useful to everyone, which means by default they're optimized for no one in particular. The good news is that's easy to fix. You don't need to learn to code or take a course. You just need to give the AI a little more to work with.
Here are five specific things you can do right now to get answers that actually feel like they were written for your business.
1. Tell it who you are
The single biggest upgrade you can make is a one-sentence description of your business at the start of every prompt. Try this experiment in whatever AI tool you use most:
First, send this exactly as written:
"Write a follow-up email to a client who hasn't responded to my proposal."
Then send this:
"I run a solo bookkeeping practice that works with small restaurants and food trucks. Write a follow-up email to a client who hasn't responded to my proposal in a week. Keep it warm but direct — I want to stay top of mind without being pushy."
The second prompt takes 20 extra seconds to write. The output is night and day. AI doesn't know your industry, your clients, or your tone unless you say so. Once you start doing this consistently, you'll wonder how you ever prompted without it.
2. Specify the format you actually want
"Give me ideas for social media posts" will get you a numbered list of ten bland suggestions. Try this instead — paste it directly into your AI tool:
"Give me three Instagram captions for a coffee shop running a Monday morning special. Casual tone, a little humor, under 150 characters each. No hashtags."
That's a prompt you can actually do something with. The more specific you are about what the output should look like — length, structure, tone, what to leave out — the better the content that comes back. Think of it less like asking a question and more like giving a creative brief.
3. Give it a role to play
AI responds well to being assigned a perspective. Instead of asking for general advice, tell it which lens to look through:
"Act as an experienced marketing consultant who specializes in working with independent retail shops. I'm a boutique owner in a mid-size city and foot traffic has been down this summer. What's the first thing you'd look at changing about how I'm promoting my store?"
Try that prompt. Then try the same question without the "act as" framing and compare. This isn't a trick — it's just telling the model which slice of its knowledge to draw from. "Marketing consultant for independent retailers" is a much smaller, more useful target than "all possible marketing advice."
4. Show it an example — ideally more than one
This is the most underused move in AI prompting, and it works remarkably well. If you've ever written a good email, a strong social post, or a proposal that landed — paste it in. More than one if you have them.
"Here are two emails I've sent to clients that got great responses. Use these as a reference for my voice and style when you write for me."
You've already done the hard creative work of figuring out what sounds like you. Sharing examples lets the AI learn from that work instead of guessing. The more samples you can give it, the better it will calibrate — one example sets a direction, two or three establishes a pattern.
5. Push back on the first answer — or ask it to push back on itself
The first response is almost never the best one. Treat it as a starting point, not a finished product. After you get an initial answer, try:
"Make it shorter and more direct."
"The tone is too formal — rewrite it like I'm texting a client I know well."
"I like the structure but the second paragraph doesn't fit my situation. I only have one employee and no marketing budget. Rewrite that section."
And here's a move most people don't know about: if you're not sure how to push back, just ask the AI to challenge its own answer.
"What's the weakest part of what you just wrote? How would you improve it?"
Or:
"Play devil's advocate on this. What am I missing or what could go wrong?"
AI tools are built to iterate. The people getting the most out of them aren't asking one question and walking away — they're having a conversation.
Getting consistently useful AI output isn't about finding the perfect prompt on Reddit or paying for a course. It's about giving the AI what it needs to do its job well: context, constraints, examples, and a little back-and-forth.
Once that becomes habit, the generic answers stop being a problem.