Something really strange has happened over the last six to eight months, and I don't think we fully appreciate how strange it is because it happened so incredibly fast. I'm not talking about some gradual technological evolution that we've watched unfold over several years. I'm talking about the way I personally work on a daily basis changing almost completely in less than a year.
At this point, I barely type anymore. I work with ChatGPT, Claude, Gemini, DeepSeek, Codex, and several other AI tools depending on what I'm trying to accomplish. But one of the biggest changes isn't simply that I'm using artificial intelligence. It's how I'm interacting with it. Increasingly, I don't sit at my computer and carefully type prompts. I put a little headset on my ear and talk to these systems like I would talk to another human being working with me.
I explain what I'm trying to accomplish. I tell them what isn't working. I describe what a customer wants. I explain what I don't like about something. I tell them what I think the solution might be, and sometimes I simply explain the problem and ask them to figure out what the solution should be. Then they go to work. In a matter of seconds or minutes, I can have research, code, analysis, writing, ideas, troubleshooting, or an entire plan sitting in front of me.
What's amazing is how quickly that became normal.
Six or eight months ago, my relationship with AI was dramatically different. There was still much more typing, copying, pasting, prompting, correcting, and manually moving information around. Today, I can have what feels much more like an actual working conversation. I can explain something naturally, interrupt, clarify myself, change direction, add context, and keep moving.
But that's not even the strangest part of what has changed.
Your AI Is Talking to My AI
The really fascinating part is what has started happening with the actual human beings I work with, because they're using AI too. Customers, developers, business associates, and other people I communicate with are increasingly sending me emails, text messages, project requirements, technical explanations, and God knows what else that have obviously been written, organized, expanded, or cleaned up by artificial intelligence.
I don't have a problem with that at all. In fact, I do exactly the same thing.
Someone might send me a massive email explaining what they want changed on a project. Maybe the original thoughts were theirs, but they put those thoughts into an AI and asked it to organize everything into a professional email. Then they send that AI-generated email to me.
What do I do with it?
I give it to my AI.
My AI reads their AI-generated email, compares it against everything else we know about the project, helps me understand what they're actually asking for, identifies possible contradictions or problems, and helps generate a response. I review that response, make sure it represents what I actually think, make whatever changes are necessary, and send it back.
Now think about what just happened.
A human gave information to an AI. That AI turned the information into an email and sent it to me. I gave that email to another AI. My AI interpreted what their AI wrote and generated a response. Then I sent that response back to the other person, who may very well take my response and put it right back into their AI.
Their AI is now interpreting my AI.
It might generate another response, which they send back to me. I put that response back into mine, and my AI analyzes what their AI said about what my AI said about what their AI originally said.
At some point, you have to stop and laugh about it.
The AIs are basically talking to each other, and we're standing in the middle supervising the conversation.
And You Know What? It Actually Works
I don't say any of this as criticism. It actually works remarkably well, and in many cases it works better than the old way. Artificial intelligence can take a rambling explanation and organize it. It can take a 2,000-word email and identify the five things that actually require a decision. It can compare a new request against months of previous requirements. It can spot contradictions, summarize conversations, turn scattered thoughts into an actionable plan, and help people communicate complicated ideas more clearly.
The crazy thing is how much of what used to consume our time has simply disappeared from the process.
Think about how we used to work. We had to decide what we wanted to say, organize our thoughts, write it ourselves, type it ourselves, spell everything correctly, proofread it, fix the grammar, edit it, format it, and finally send it. Spell-check eventually helped with part of that. Grammar software helped with another part. Predictive text started suggesting words and sentences. Then artificial intelligence started helping us write.
Now I can put a headset on my ear, explain what I'm thinking in ordinary conversational English, and have a machine turn those thoughts into something useful.
And this transformation didn't take ten years.
For me, the most dramatic part of it happened in about six to eight months.
Somebody Still Has to Be the Boss
There is, however, a major problem with all of this if we're not careful. The fact that artificial intelligence can do so much of the work does not mean we no longer need to understand the work. In fact, I think the opposite is becoming true. The more capable these systems become, the more important it is that somebody involved actually understands the objective.
You still have to be the architect, the engineer, and the boss.
AI is extraordinarily good at taking a direction and running with it. The problem is that somebody still has to determine whether it's running in the right direction. If nobody is paying attention, the AI-to-AI loop can become almost comical. Their AI misunderstands something. My AI responds intelligently to that misunderstanding. Their AI analyzes my AI's perfectly reasonable response to the original misunderstanding. My AI then responds intelligently to their AI's interpretation of my AI's response.
Pretty soon, two incredibly sophisticated artificial intelligence systems can be having a beautifully written, perfectly formatted, extremely professional conversation about something neither human actually wanted in the first place.
That's when a human has to interrupt.
You have to say, "Stop. That's not what we're doing. Here's what I actually think. Here's what we're actually trying to accomplish. Here's the decision, and here's what we're doing next."
That human intervention is what gives the entire system direction. AI can process the information, organize it, analyze it, communicate it, and even execute enormous portions of the work. But somebody still has to understand why we're doing it in the first place.
The Skill Has Moved Up a Level
I don't think the future belongs to people who simply know how to "use AI." Before long, saying you know how to use AI is probably going to sound about as impressive as saying you know how to use Google or send an email. Of course you do. Everybody will.
The valuable skill is going to be knowing how to manage AI. You need to know which systems are good at which jobs. You need to know how to provide the right context. You need to recognize when an answer is wrong even when it sounds extremely convincing. You need to know when to let the AI keep working, when to redirect it, and when to stop it completely.
Most importantly, you need to know what you're actually trying to accomplish before you ask artificial intelligence to accomplish it.
That's the part I don't think AI eliminates. If anything, AI makes that skill more important because when the cost of creating something approaches zero and the time required to create it gets shorter and shorter, the bottleneck is no longer typing. It isn't spelling. It isn't formatting. In many situations, it isn't even writing or coding anymore.
The bottleneck becomes judgment.
What should we build? What should we say? Which idea is right? Which answer actually solves the problem? What matters and what doesn't? When should we keep going? When should we stop? When is the AI confidently heading down the wrong road?
Those aren't typing decisions. They're management decisions.
We're Becoming Managers of Machines
That's probably the biggest change I've noticed in my own work over these last six to eight months. I'm still creating things. I'm still solving problems. I'm still making decisions. In fact, I may be doing more of those things than I ever have before. But I'm physically doing less of the mechanical work and spending far more of my time directing the work.
I'm not spending nearly as much time typing the thing. I'm spending my time deciding what the thing should be. I'm not manually writing every line of code. I'm deciding what the software should do, determining whether it actually does it, identifying what needs to change, and directing the AI to make those changes. I'm not writing every email from a blank screen. I'm determining what I actually want to communicate and making sure the AI communicates it accurately.
That's a very different job than the one digital creators had even eight months ago.
I don't believe we've reached the point where humans are unnecessary. I think we've reached something much more interesting: the human role is moving up a level.
The computer can type. Let it type. The computer can spell. Let it spell. It can write, research, code, analyze, summarize, organize, and respond to another computer. Fine. Let it do all of that too.
Our job is becoming something different. We have to understand the objective, provide the context, recognize the nonsense, make the decisions, and interrupt the loop when necessary. We have to manage not only our own artificial intelligence but, increasingly, conversations in which somebody else's AI is communicating with ours.
In other words, we still have to be the architect, the engineer, and the boss.
Because no matter how intelligent the machines become, somebody still has to know what the hell we're actually trying to accomplish.
