Most People Use AI Like Software. That's the Problem - E66

Speaker A: In episode 65, you turned on the black light and you saw it. The opportunity that was already on your page. The work that could not be done before at any price, sitting right there in your own business. We spent the whole episode inside that first room. And so you did the brave thing. You walked into the second room, you opened up the tool, you pointed it at the opportunity you finally saw, and you hit a wall that nobody warned you about. Because seeing the opportunity is not the same as leading the thing that chases it. You typed out what you wanted and the output came back fine. Not great. So you tried a different tool, same result. And a quiet thought crept into your mind. Maybe this stuff is just not ready. No, here is what actually happened. You walked into that second room holding the most powerful tool any leader has ever held, and you ran it like software. You gave it a command and you hit go. The move was never to command it. The move was to lead it, to manage it like a person. You are responsible for developing. This is leadership unlocked. Let me describe what this looks like at your desk on a Monday. You tell me if this is you. You sit down with the tool open. You type a sentence. Do this for M. Me. Build this, clean this up. You hit go. A few seconds later, it hands you something. It's fine. It is not wrong. It's just not what you would have done. So you judge the output. You feel a flicker of disappointment, and you close the window. Maybe the tool is overhyped. Maybe you need a better one. Now, let's be clear. This is not a tool problem. This is a posture problem. You talk to it the way you would type into a search bar. One shot, exact words, fire and forget. You did not lead it the way that you would lead. A new hire, an incredible leader I work with named this without meaning to. She runs the finance side of a growing company, and she put her whole struggle into one sentence. What should I be asking the tool to do? And how should I be asking it? And that's the real question. Not which model, not which subscription, not which prompt engineering thing should I be using. Exactly what do I say and how do I say it? And here's the answer almost nobody gives you. You were trained to treat this like a Google search, where you tell it exactly what to do. Not anymore. The most capable thing you will ever delegate to does not want a command. It wants direction. The cost of getting that wrong shows up in three places. It shows up in your team who get the same flat output that you got. Decide that AI is not ready for the real work. And they go back to the old way. It shows up on your plate because the work AI could have carried stays where it was. And the headcount math keeps tightening while you keep doing it yourself. And it shows up in your future because you build no muscle for this. No muscle memory, no expansion of capability. And so when the next tool lands and it will, you start over from zero. Now, uh, here's the part that should make you angry in a useful way. The standard fix is wrong. Everyone treats the gap as technical. Take a prompt engineering course. Learn how the model works up under the hood. And none of that touches the actual gap. The actual gap is that you already know how to lead a person to a great outcome. You have done it for years and used almost none of that skill. The moment you sit down in front of the machine, you give it orders. You never give it a charter. You never tell it what great looks like. You never ask what it would propose or what questions it has before it starts. Everything you know about getting the best out of a human, you leave at the door. So here is the permission that I want to give to you, and it's the same kind that I gave you over the last two episodes. That disappointment that you felt staring at that flat output is not proof that you're behind. It's your judgment telling you that there's something off here, that there's not quite right about what we're doing. And you're right. You've been pointing one of the most powerful instruments that any leader has ever held at the problem in this second room, the one that you walked into from that first space. And you're running it like software. Let me tell you about the morning I watched two incredibly sharp leaders catch themselves doing exactly the same thing. A few weeks back, I got on a call with two amazing leaders that I work with. Uh, Jamie runs a construction and engineering firm, and she sits on multiple boards, and Tricia runs the finance side of that same business. Both of them manage people for a living. Both of them are good at it. And both of them sat down to use AI Uh, and felt exactly what you have felt. Tricia said it very plainly. Her struggle was not the software. It was knowing what to ask it to do and how to ask. She kept defaulting to things that she already knew how to automate another way, because that was the only language she had for talking to the tool Jamie named the deeper version. She said the honest reason that work piles up is that it is just easier to do it myself than to teach somebody else how to do it. And then she said the thing that turned that around. You have to spend the time teaching it, but once you do, the outcome is better. And sit with that for just a minute, because she was talking about people. And it is the most precise description of managing AI I, uh, have heard from anyone. It feels slower to teach than to do. And so you do it yourself and you stay buried. Let's think about this for a minute. That has been the number one challenge for delegation for decades. It's easier for me to do it, so I'm not going to delegate. And when we think about it, that's exactly the problem we're facing with AI. Now, let me be honest about my own side of this, because the honesty is the whole point. I did not arrive at this in a clean moment of inspiration or with a clean track record. For a long stretch, I was the leader giving the machine orders, go, do this, build that. And it would come back with a flat response. I was blaming the tool in. And then I would reach for a different one. I was managing the most capable thing, just like I was managing a vending machine. Put in a command, get out a snack for me. The turn came in a strange moment. I was showing Jamie and Trisha how I actually work now live. And on that call, I had built a set of assistants. And each one is set up for a very different job. I went to hand one of them a task, and I couldn't remember its name. And I said out loud, I can't remember the name of this one. I need to rename that thing. And the second it left my mouth, it hit me. I had just forgotten the name of one of my workers, the way a CEO forgets the name of an account executive who started three weeks ago. My head exploded a little right there on the call because I was not using software. I was running a team. I was giving a team of employees clear direction, making sure they knew what great looked like, coaching them when they missed. And it wasn't prompting. That's people management. I've been doing the most ordinary thing a leader does and doing it with workers made of software. And here is why I'm telling you about Jamie and Tricia instead of just telling you about me. They are not my students in this story. They are you two senior operators, excellent at leading people who sat down at the tool and hit the same wall that I see everyone hitting the same wall that you might be hitting. The shift that happened on that call when this analog moment happened when. When you could see them realize, wait, I already know how to do this. That's the shift I want for you. Not because I taught them a new skill. I didn't. Because they recognized one they already had and they saw where to point it. This is the frame I lead with now, and it's the one that I want to hand to you. Because I've recognized that over the last year, year and a half, even two years, I have been treating the machine like a team. I've just been leading the way I lead people. And it creates different and better results when I do that. It's been an evolution. I've been learning this and getting better at this. And it's in contradiction to what I'm seeing right now. Most leaders are running A.I. uh, in what I call software mode. The discipline I want you to adopt is to run it in people mode. It's the same tool. It's a completely different posture. The difference is not technical. It's the difference between giving a command and giving direction. Last week we talked about how the right question helps a person see the dots on the page that our job as leaders is to shine the black light that lets them see. This is that same instinct turned towards a tool. You're not operating it, you're developing it. Let me walk you through the five things that you already do well with a great employee and what you each one looks like in software mode versus people mode. And you're about to recognize every single one of these because you already do them. Number one, expectations. In software mode, you type in a one dimensional line and you hit go build a forecast. That's the whole brief. In people mode, you write the charter first. You would never hand a new hire a task with no context and no sense of the priorities and no alignment to where we're going. And you should do it with a tool either. I say it to myself like this. I am creating the job description for an employee. Here's the context, here's who you are. Here are your core responsibilities. If you write it once every command after it lands on a foundation instead of just a new command prompt. So start with clear expectations. Start with the Charter 2, a system of measure. In software mode, there is no measure. You eyeball whatever comes back and decide if you like it. In people mode, you tell it how the work will be judged before it starts. You would tell a person, hey, here is what good looks like and here is how you will know you're hitting the mark. So you should tell the tool the same thing. Here is What I will be checking here is what I will be expecting back from you from a measure perspective. If you set the standard up front instead of reacting to whatever it shows you, you get a better result. 3. A definition of done. In software mode, you say go clean this with no picture of what clean means. And then you're surprised if it doesn't give you exactly what you want, if it doesn't give you the picture that you have in your mind. In people mode, you define what great looks like first. This is the oldest move in your leadership toolkit. What is the goal? What is the desired outcome? What is it we're trying to achieve? What does done look like? If you get that wrong with a person, you get busy work, you get a second, uh, iteration, you get a missed target. If you get that wrong with the tool, you just get the same thing but much, much faster. Number four, coach and train. In software mode, you take one shot, judge it, blame the tool when it misses, give it another response, and eventually maybe you get on target. And here's the truth every leader knows and keeps forgetting. It is slower the first time to teach than to do. I watched a CEO I work with live out this exact lesson with his human successor. He was very intentional about how he stepped back back from the work. He would give feedback and then let the other person figure it out because he knew that if he stepped in to solve the problem, he would never actually build up someone who could. And it's the same move with the tools that we leave. If you show it how you think, if you correct it, if you build it up over a few reps, it then carries the work. The leader who refuses to spend the time teaching, training, evolving, enhancing. The tool never builds the capability, not in a person, in a tool. Number five, a culture of accountability. In software mode, you turn it loose, let it run unsupervised, and blame it when the output is off. In people mode, you build the gate. Nothing ships without a check. I don't let my go do things before it confirms what it is going to do or ask me its questions or proposes the approach first. That's not micromanagement. That is the same accountability that you build into any team. Clear authority, a checkpoint before action, a review before it ships. Now here is the move that ties all five of these together and it lives in a single set of keystrokes. In software mode, you give the command. In people mode, you ask the question the next time you go to type something into your AI, uh, tool set. Whichever one that you use do this for me. Stop and type four words instead. What would you propose? For example, here is my problem. What would you propose to solve it? That one change pulls expectations, definition of done, and accountability into the same breath. Because now the tool has time to think, so to speak, before it acts and you react to a plan instead of cleaning up a result. I have learned that if I tell it what to do, it will do it, but it will never bring to bear the possibility it could have created. Ask instead of tell, and you get that possibility again. Think about this through the lens of leading people. If you tell people what to do, you are limiting them. You are giving them a ceiling. But if you ask them what they think, if you ask them how they may solve the problem, you're raising their floor. We want to do the same thing with the tools that we lead. This simple mindset shift allows you to ask better questions. And then just one more question that goes right behind that. What would you propose? What questions do you have for me before you proceed? That's the question you would ask any sharp new hire before they run with something. It surfaces the assumptions that you did not know you were making. It is people management in the prompt window, and you will know that the analog is real the day it happens to you the way it happened to me. The day you reach to hand off a task and you cannot remember the name of the worker you're handing it to, that's not a glitch. That's the tell. You forget an agent's name the way you forget that new hire's name. Because at that point, you were not operating software. You're running a team. And then the real magic happens is you no longer forget the names because you really lead like a team. A few episodes ago, I made the case that your leadership matters more in this era, not less, and that the principles do not expire when the tools change. This is what I meant. Clarity, alignment, execution. The whole value creation chain, the whole leadership execution chain still holds up. Uh, you're just running it on a worker made of software now. So here's what I want you to do this week. It costs you nothing but to avoid the impulse to give an order, and you can run it in your very next AI session, take the next prompt you are about to type, and before you type it, do two things. One, write it as a question. Replace. Do this for me with. Here is the problem I'm trying to solve. What would you propose? You are not handing it a command, you are handing it a problem and asking it to give you perspective, input, what it thinks. 2. Before you turn it loose to start doing the work, ask the second question. What questions do you have for me before you proceed? And then read what comes back. It will ask you for any context that you did not give it. It will surface an assumption that you didn't know that you were making. It will offer an option you would not have considered on your own. Now, I'm going to be honest with you about the catch here. The first two days, this will feel slower. It will feel like you're explaining yourself to a machine. We could have just done the task. And that feeling is real and is the same tax you pay when you're coaching a new hire instead of doing the work yourself. It's the teaching tax and it pays back the same way that it always has. By Friday, your prompts look a little different. Your first pass results are a little sharper. By the end of the month, the work that is being done is the work that you use to carry yourself. And here's the part that really compounds by then, your team's prompts are starting to look different too, because they've watched how you lead the tool, you've shown them what your getting that's better and you're sharing these ideas with them and they begin to lead the same way. Do this for a week. Every meaningful AI interaction. Stop telling. Start asking. Let the worker think before it acts. You already know how to do this. You do it with people every day. You were just doing it in a new room, now with new team members. So let me leave you with the whole thing in one breath. You do not need a new leadership skill. You need to use the one you have already spent an entire career building. Stop typing commands into software. Start giving direction to a team you cannot see. Write the charter, define what great looks like. Coach it across reps, hold it accountable. Uh, and when you go to tell it what to do, ask it what it would propose instead. And here's the one line that I want you to carry out of this. Do not manage AI uh like software that you fire commands at. Manage it like a person. You are developing. You already know how. You just haven't pointed that skill at the machine yet. Everything that we've been talking about over these last few episodes, the lever, the black light, managing the machine like a person, it's all based on one idea. Our ability to lead others is a based on our ability to ask the right questions, to cast the right vision, to ensure that people understand where we're going and where they fit in. These concepts aren't new, they are ancient principles. Those principles are what will create the difference for the leader in the age of AI uh. There's much more to come on this topic because it is something that I'm hearing so much about and the more noise we have in our system about tools and technology, the more important leadership is. AI uh will either amplify the impacts of great leaders or it will expose the flaws of those who are not leading. Hey, before you go, if this episode helped you see leadership differently and I have three personal requests. First, and this is the most important one, share this with one leader who needs it. Think of someone in your network who's frustrated with their team's execution or is trying to scale their own impact. Copy this episode link right now and text them, hey, I just listened to this episode and I think you'll get real value from it. That simple act of sharing that is servant leadership in action. Second, subscribe to Leadership Unlocked. Whether you're on Apple Podcasts, Spotify, YouTube, or your favorite podcast player, just hit subscribe or the Follow button. That way you won't miss the next episode and it helps us reach more leaders like you. And third, if you're willing, this will be a huge help for us. Leave a five star rating and write a quick review. I know it takes 60 seconds of your time, but here's what it does. It helps other leaders discover the show. Every rating, every review. It's how we expand our impact together. So those three things Share with a friend. Subscribe, Leave a review. I'm deeply appreciative and that would give us so much help. Thanks for being part of the Leadership Dunlop community. Now, um, go create some clarity for others.