Share

Read the author’s previous article on AI, HERE and HERE.

I’ve never had success asking AI to simply write an article. AI can write competent prose, but it doesn’t sound like me. I write with a certain rhythm, and my sentence construction and word choice are often unconventional. I use ellipses and em-dashes not as punctuation but as lyrical emphasis. I prefer to italicize words rather than sandwich them between quotation marks. I believe colons and semi-colons should be banned from all modern prose, and the word that should be struck from the English language.

I didn’t use AI to write the article. I used it to accelerate insight.

Instead, I use AI as a thinking enhancement tool. In the case of this commentary, I used it to explore possibilities, challenge assumptions, and test ideas. Somewhere during the process the specific point of this article became clear. Would I have discovered it without AI? Almost certainly, but I would have wrestled and sweated over it for far longer. I didn’t use AI to write the article. I used it to accelerate insight.

When most people sit down to use artificial intelligence, the first thing they do is start typing. It seems perfectly reasonable. After all, AI is a conversational tool, and conversations begin with words. It was certainly my approach when I first started experimenting with it. I’d open a chat window, ask a question, and wait to see what came back. Sometimes the results were impressive. Other times they missed the point completely.

This tool answers back.

The real problem wasn’t AI. It was me. I wasn’t clear about the problems I was asking AI to solve—the old saw of garbage in, garbage out. I was expecting a science fiction thingamajig one step away from sentient. After all, it’s what the AI fearmongers are constantly pushing. But what I was encountering was a sophisticated tool I first needed to learn how to use. The only difference is this tool answers back.

AI doesn’t think for me. It gives me something to think about.

The fact AI answers in complete sentences makes the interaction feel conversational, but appearances can be deceptive. I’m not having a conversation with another mind. I’m using AI as a thinking enhancement tool to explore my own ideas. Every response becomes something I can evaluate. Sometimes I agree. Sometimes I disagree. More often, an AI suggestion nudges my thinking in a direction I hadn’t considered. AI isn’t doing my thinking. It’s helping me think more effectively. More precisely, AI doesn’t think for me. It gives me something to think about.

Better conversations lead to better conclusions.

This process has changed the way I think about AI. The conversation with AI may begin with the first question, but the thinking shouldn’t. By the time I open a chat window, I’ve usually spent quite a while thinking about the problem. Who am I writing for? What am I really trying to say? Where am I uncertain? What do I already know, and what am I hoping to discover? Those questions don’t produce finished prose, but they lead to better conversations with AI. And better conversations lead to better conclusions.

When I present training seminars to law enforcement, one of the things I emphasize is the success or failure of an interrogation is determined before the first question is asked. Almost always this statement is met with blank stares, but by the end of class the meaning becomes clear. The axiom is true not because the first question doesn’t matter. It’s because the questioning is only one step in a much larger process. The thinking, the preparation, and the objective should all be in place long before anyone begins asking questions. Working with AI has reinforced the same lesson but in a different setting.

Until you know what you want, you’re asking AI to solve a problem you haven’t defined.

The biggest change in my working relationship with AI has been learning to separate the assignment from the problem. The assignment is what you’re doing. The problem is what you’re trying to solve. For example, the assignment might be to write an article, prepare the framework for a lesson, or outline a series proposal. But the problem is deciding what you’re really trying to communicate. Until you know what you want, you’re asking AI to solve a problem you haven’t defined.

Article One in this series is a good example. We didn’t begin by writing about artificial intelligence. We began by deciding what kind of article we didn’t want to write. It wasn’t going to be a technical manual. It wasn’t going to be a warning about machines taking over the world. It wasn’t going to be another list of prompts guaranteed to make life easier. Only after eliminating those possibilities did we discover the article was really an invitation to explore a new tool together. Once we understood that, the writing itself became surprisingly straightforward.

This experience taught me another lesson. AI is remarkably good at helping me develop an idea, but it isn’t particularly good at giving me one. The idea still has to come from somewhere. It comes from experience, curiosity, observation, or sometimes a conversation over breakfast. Once the idea exists, AI becomes remarkably effective at helping with development. It asks questions I hadn’t considered. It exposes weaknesses in my reasoning. It nudges me toward possibilities I might otherwise have missed. None of those things create the original idea. They help me develop it.

The effort of putting my thinking into words often clarifies it for me.

I’ve also stopped expecting AI to solve every problem. Sometimes I don’t need an answer. I need help identifying the question. Other times I need someone—or something—to push back against my assumptions. There are also times I’ve rejected every suggestion AI offered and still considered the exercise a success. Why? Because the effort of putting my thinking into words often clarifies it for me.

I now spend much less time worrying about prompts than many AI discussions suggest. Good prompts are useful, but they aren’t magic. They grow naturally out of understanding the problem. If I’ve laid the groundwork, the prompt almost writes itself. If I just jump in, my confusion becomes the garbage going in. Again, in that situation, the problem is not AI. It’s me.

Ironically, I’ve found the preparation is usually more valuable to me than it is to AI. The computer doesn’t care whether I’ve organized my thoughts. I do. By the time I’ve gathered my notes, considered my audience, and decided what I’m really trying to accomplish, I’ve often untangled half the problem before I’ve typed the first prompt. Occasionally I’ve even solved the problem without opening AI at all. I don’t consider this a failure because it means I’ve solved the hardest part of the problem.

I don’t expect—or want—AI to think for me. Instead, I use it to sharpen my thinking.

The real value of AI, at least for me, isn’t finding instant answers. It’s what happens after I’ve done enough thinking to ask better questions. That’s a very different way of looking at the technology. I don’t expect—or want—AI to think for me. Instead, I use it to sharpen my thinking.

If you’re experimenting with AI, try one small exercise before opening your next chat. Don’t start by typing a prompt. Spend a few minutes defining the problem you’re trying to solve. Not the assignment—the problem. Decide what you’re trying to accomplish before the conversation begins. You may discover what I did—the quality of the conversation depends less on the wording of the first prompt than on thinking before you begin typing.