One of the most frustrating experiences for me as a parent is asking my kids why they did something and them responding, “I don’t know.” There are a number of reasons they might say it.

Maybe they think they’ll get in trouble if they tell me why they did it. Maybe they did it to protect a sibling from getting in trouble. Maybe they don’t think, “because I wanted to,” is a good enough reason.

But it’s mostly because kids have no impulse control. There’s very little friction to go from “I should do this” and actually doing it. And there’s even less thinking.

Over the last couple of years, I’ve noticed this same trend with using AI. AI apps are getting more powerful, easier to use, and literally everywhere¹. There’s no friction to use them. And there’s potentially even less thinking.

Why We Use LLMs Too Much

About a year ago, my wife provided me with two examples of AI use: one was super good, and one was not.

The super good use was designing my birthday cake. She described what she wanted, and then gave the design to an incredible local baker to actually make. She had an idea, but couldn’t execute it in a way that would clearly communicate to the baker what she wanted.

The not so good one was using it to come up with a scavenger hunt for our kids at the local trail. When she told me, I responded (bluntly), “You couldn’t come up with 10 things for them to find?”

Now look — I get that she’s busy and we had a lot to do that morning and it was going to be a fun activity for our kids. It was, as you could imagine, a pretty good list. But that’s not the point. The point is that she didn’t even think about not using it. It was easy and it was a task it could do.

I fell into this trap a few months before that; I was talking to it on my phone, asking it questions instead of thinking about them myself. But instead of creating a benign game for my kids, I ended up taking my business in the wrong direction.

I, like so many other people past and present, was taking the easy way out; but I was also tolerating substandard work. Luckily, I realized before it was too late.

When we’re not conscientious about how we’re using AI, we risk using it for important stuff (like critical thinking, articulating opinions, or work that requires taste and judgment). And then everything we do becomes substandard.

So how do we make sure we’re thinking about our AI use?

What is Conscientious AI Use?

The shortest possible answer to this question is this: You don’t blindly use AI to do stuff. And the goal of using AI should go far beyond, “I don’t have to do it anymore.”

But it’s a little bit deeper than that. Earlier this year I was having a conversation with a friend about how it seems people are getting “AI-Brained.” They struggled to function without using AI to help them.

This led me to write 4 warning signs you’ve got AI Brain Rot, and I think it’s one of my most important blog posts to date. I would strongly recommend you read the whole thing, but the 4 signs are:

1. You use AI reflexively without thinking through a problem yourself first.
2. You skip talking to people and instead ask AI questions humans should be answering.
3. You have no skin in the game because you’re just blindly asking AI to do things.
4. You treat AI as an advisor or coach.

In that article I go on to explain why these are very bad paths to go down, and how you can prevent AI Brain Rot. I even built a quiz to help you figure out which area you are highest risk for². But something missing from that article is the other side: a set of principles for using AI conscientiously³.

How to Use AI Conscientiously

There are 3 rules for using AI conscientiously:

#1: AI is for tasks, not craft. I will die on this hill. If you’re using AI to wholly create something for you, you are not using it conscientiously. And I’d argue it’s not really yours.

There is a nuanced difference in using it to assist you in creating something. But it’s not the same as something like Photoshop because Photoshop will not try to predict what you want it to do.

Instead, we should use it to take repetitive tasks off of our plate, or do things that have clear and verifiable outcomes. Some examples are:

1. Move tasks from our inbox to our task app
2. Do verifiable research on a topic (as in, cite sources that we can then check)
3. Write scripts to connect to other services (the scripts don’t change when the models change).

#2: AI cannot replace judgment. Asking it for “advice” and seeing what it comes up with betrays conscientious use because it means we’ve forgotten that it’s not actually judging anything — it’s just giving you a statistically likely answer.

As the models get more impressive, it’s easy to think that has shifted, but it has not. It will make a calculation that the answer it’s spitting back at you “looks” like some advice a human would give; but if you’re asking it judgment questions, it’s probably not. Just ask the hikers who had to be rescued after using AI to plan their hike.

#3: There should always be a goal for AI use beyond “I don’t have to do it anymore”. This is really the core tenet of Conscientious AI Use. I put it 3rd on the list because we need to remember 1 and 2 before we even get here.

But once we do, we can keep in mind the best uses for AI and then determine, based on our goals, if we should use AI for the task. I have more here, but I developed a 3-question test for this exact purpose:

1. Does this task actually need to be performed? If it doesn’t, then neither you, nor AI, should do it.
2. If it needs to be performed, ask, “what is the goal of the task?” Without knowing the goal, you cannot possibly know if AI is actually helping you with it.
3. Then finally, “how can I measure if and how AI is helping me with this task?”

Understanding that AI is for tasks, not craft, that it doesn’t have any ability for judgment, and that we should have a clear reason for using AI combines to create Conscientious AI Use.

Who is Conscientious AI Use For?

I’d love to say Conscientious AI Use is for everyone. I firmly believe it should be for everyone.

But the truth is if you think AI is the end-all-be-all, that you have a “team” of “agents” “doing” your work for you, that you refer to AI as people and that it is your employee, I’m not going to convince you otherwise.

I refer to this as being AI Blue Pilled, and if people want to keep believing that AI is capable of thinking for them, I’m not going to change their hearts and minds until they choose to unplug from the proverbial Matrix⁴.

But, if you’ve gotten this far, I suspect you are not AI Blue Pilled. If you know AI can help you (because it can), you’re just not sure, you’re in the right place.

If you’re worried about becoming overreliant on AI to the point where you’re not thinking for yourself, this is for you.

This will serve as your guide for thinking about how you use AI, and what you use it for.

That way, when someone asks you why you used AI for the task, or if using AI actually helps you accomplish your goal, you’ll have a better answer than, “I don’t know.”