Why most people get poor results from AI, how a TEDx talk on third-grade reading explains what's actually happening, and three things you can do this week to fix it.
Every time I got a bad response from an AI tool, I went back and read what I had typed. And almost every time, the problem was what I wrote — not what the AI generated. The AI was doing exactly what I asked. I just wasn't asking clearly enough. That realization stopped me cold. If I — someone who has been working in technology for years and currently builds and tests these tools — was struggling to communicate clearly with AI, what does that say about the rest of us?
Here's the uncomfortable answer: it says a lot. And the data backs it up. As of 2024, 54% of American adults read below a 6th-grade level. Between 2017 and 2023, adult literacy scores dropped 12 points — the largest recorded decline in the history of that assessment. This is not a coincidence at a time when AI tools are being deployed across every industry. The two problems are connected, and once you see the connection, you cannot unsee it.
Vague Instructions Are a Productivity Tax
If you're using AI in your business or at work — and most of you are or are about to — every vague instruction is wasted time. You prompt, you get something unusable. You prompt again, you get closer. You prompt a third time. That loop kills the productivity gain AI was supposed to give you.
The math is straightforward. If AI should cut your time on a task by 40%, but you spend that time re-prompting because your instructions weren't clear, you've gained nothing. You've just moved the work from doing the task to arguing with the tool about what the task is and what it should look like.
Bottom line: unclear communication is not a soft skill issue. For your business, right now, it's a productivity tax.
AI Is Not a Search Engine. Stop Treating It Like One.
Most people treat AI like a search engine — type a few words and expect it to figure out what you meant. That's not how it works. AI is more like a new hire on day one. You cannot hand a new employee a task on a sticky note and expect a finished product. You have onboarding for a reason. The same applies here.
You have to tell AI what you need, why you need it, what format the result should be in, what a good result looks like, what to avoid, and who is going to use it. That last one matters more than most people think. The person reading your output is not you — and AI needs to know who it's writing for. The more clearly you explain the job, the better the output. That has nothing to do with AI. That's basic communication.
What Happens When You Don't Give Enough Context
I run an online store selling landscape photography from my trips — prints on canvas and acrylic. When I first used AI to write product descriptions, I would give it the location and the name of the shot. Peru. Italy. What I got back was a travel brochure. Factually accurate. Completely useless for selling a print.
The AI had no idea I was selling a physical object. It didn't know the buyer was standing in their living room trying to picture this on their wall. It didn't know the texture of an acrylic print changes how light hits the image. It didn't know the emotional response I was trying to create. I hadn't told it any of that. Once I changed the instruction — "write a product description for a 24x36 acrylic print of the Costa Verde in Lima, Peru, for a buyer decorating a modern living room, focused on how the image makes them feel standing in front of it" — the output was completely different. Same AI. Same tool. Completely different instruction. Completely different result.
Bottom line: AI amplifies what you bring to it. Bring clarity, you get clarity. Bring a vague idea, you get a vague answer. And we are less clear than we think we are.
AI Can Be Confident and Wrong at the Same Time
Here's the part nobody talks about. When I was debugging an automation workflow — a series of connected steps that had stopped working — I gave Claude the exact error output, copy-pasted directly from the system. Precise information. Clear context. Exactly what you're supposed to do. Claude gave me a confident answer: fix this node, here's how.
The fix didn't work. Because the problem wasn't that node. It was the node before it — the one that wasn't passing the right data downstream. Claude had diagnosed the symptom, not the cause. I only caught it because I understood enough about how that workflow was built to know something wasn't adding up. If I had trusted the output without understanding the system, I would have spent hours fixing something that wasn't broken.
This is why the human in the loop is not optional. The value you bring is not just asking better questions. It's knowing enough to evaluate the answers. That verification step requires more expertise than the generation step — not less. The point of using AI in your business is not to hand over your work and walk away. It's to do more, faster, with you still in the room — because now you have the time to do other things you didn't have the bandwidth to do before. The moment you remove yourself from the equation entirely, you have transferred your professional judgment to a tool that doesn't have any. And you still sign your name to the output.
Bottom line: the human in the loop is not a formality. It's where the actual value lives.
Why This Is Happening — The Literacy Connection
Last weekend I was at a TEDx event in Boca Raton. The presenter made a point that stuck with me: through third grade, you learn to read. After third grade, you read to learn. That transition — from decoding words to processing and organizing ideas — is the same cognitive skill that lets you write a clear email, a clear brief, a clear prompt. And critically, it's the same skill that lets you read an AI response and evaluate whether it actually makes sense. They're not different skills. They're the same skill applied in different directions.
Between 2017 and 2023, adult literacy scores in the US dropped 12 points — the largest recorded decline in the history of that assessment. The people who struggle most with AI tools are often not struggling because the tools are hard. They're struggling because writing a clear instruction and then evaluating whether the response is correct requires a mental process that is weaker across the board than it was six years ago. That is sitting inside your workforce right now.
Bottom line: the productivity gap AI is supposed to close has a floor. That floor is built on whether your people can communicate clearly and verify what comes back. If that foundation is soft, AI won't fix it. It will just make the gap visible faster.
Three Things You Can Do This Week
First, treat your first AI output as a draft, not a result. The goal of your first prompt is to find out what information the AI needed that you didn't give it. Read what came back, figure out what's wrong or missing, and add that to your next instruction. One thing I do at the end of a first prompt: "Ask me any questions you need answered in order to complete this task." That alone changes the quality of what comes back.
Second, before you prompt, write one sentence describing what a good result looks like. Not what you're asking for — what the finished thing should accomplish, who will use it, and what they should do or understand after they engage with it. If you can't write that sentence, the AI can't produce the output. You haven't figured out what you actually want yet. If you don't know, the AI doesn't know.
Third, never accept a confident AI answer on something you cannot verify. If AI gives you a fix, a recommendation, or a conclusion and you don't understand the system well enough to check whether it's right, that's your signal to slow down — not to trust it faster. The verification step is not optional. It is the job.
None of these are AI skills. At the bottom, they are communication and critical thinking skills. The AI just makes the cost of skipping them impossible to ignore.
The Bar for Communication Just Got Higher
AI is not lowering the bar for communication. It's raising it. Because now your communication has consequences you can see in real time. A bad instruction produces a bad output immediately — no social filter, no one politely filling in what you probably meant. Most people experience that as frustration with the tool. The people who get ahead with AI experience it as feedback.
The ability to communicate clearly and evaluate what comes back is a skill. It can be built and it can be practiced. But it has to be treated as the actual work, not as something AI handles for you. The question is whether you treat the next bad AI output as a reason to blame the tool, or as information about what you need to do differently.
If you want help applying this inside your business — mapping where AI fits, where the risks are, and how to build workflows your team can actually verify — book a free call here. We'll look at what you're working with and figure out where to start.
