Some of you may know that I have been ballroom dancing competitively for about five years. (That story is a bit random and happenstance: think post-pandemic, half-marathon burnout, random walk in Marylebone.) At the time of this writing, I’m headed to the Blackpool Dance Festival to compete in the ProAm event (the professional contest celebrates the 100th anniversary this year). I also have been loving ChatGPT. So…naturally I decided to ask AI about new routines for the competition. I spent some time working through what I had done in the past, what interested me. I’m not really a novice anymore. Then I confidently took it to my partner and teacher, who has been dancing since she was five. And what I learned was a good lesson for anyone who wants to be a leader, including next gen leaders trying to improve.
The real limitation of AI, for next gen leaders, isn’t in the tech. It’s in the mirror.
My very patient and now quietly smirking partner nodded as she read through the very thorough and well-articulated AI description of the routine. I sort of sensed this wasn’t going to end well so I asked immediately if the steps were real. Yes, she confirmed, though we would need a choreographer to really put it together. But she wanted to show me the steps themselves. She started dancing…and I started realising that this was not what I’d imagined.
First problem: I wasn’t sure I could do all the things that apparently were on this genius doc I’d brought. Worse, I didn’t want to do all the things — it just didn’t look good. Certainly not competition-level good.
There was nothing wrong with the output in theory. But my dance partner also knew, on sight, without dancing it, that it wasn’t a great option. That isn’t a disastrous discovery — someone with decades of experience realised that I knew enough to be dangerous. And not much more.
This is in fact what I think might be an issue for next generation leaders who are relying on AI increasingly, right when they might need more outside coaching. They too have enough information to be dangerous. I felt fully competent enough to ask AI about ballroom dancing. And I clearly did know enough to produce a real, viable option. But usually we are looking for the best possible option. Certainly for a competition, and definitely when it comes to career decisions. When you know nothing it’s easier to realise that you don’t have the basic data necessary to even ask AI how to solve a problem. When you know a lot, you feel confident that you know enough. And that’s the problem — knowing enough to feel confident lives terribly close to not knowing enough to realise the gap in your knowledge.
Experience isn’t linear, usually. It compounds. Someone who cooks regularly can’t just execute a dish — they understand what each ingredient is doing, and why this spice pairs well with that protein. What can you substitute because it has the same effect on the cooking time. Dance is the same — that step works but it’s physically more difficult, without any payoff artistically. You can move from A to B technically, but to do so is awkward and breaks the momentum. I would say she knew instinctively, but I think instinctive is at risk of a bad rap in the era of AI. After all, I brought facts. But she brought experience.
Her instincts are built on years of learning, practicing and observing. She knows why those steps are not common…and it wasn’t because I was a secret dance genius uncovering a never imagined new routine. Leadership is the same. It’s not one thing, usually, it’s eight. Or 800. It isn’t that the message is wrong, but the timing is awful. The problem isn’t that the person can’t lead technically, it’s that the team in question is understaffed and has been in place less than a year. I can easily imagine a CHRO looking at an org chart proposal written by a leader who knows a very good amount about managing people. And not enough to know the plan is problematic.
All of this, of course, goes beyond just AI. We are all victims to this phenomenon: you stop being fully ignorant of necessary skills and you start feeling like you’ve actually mastered them. Relatively speaking, I know so much more about ballroom dancing than I could have ever imagined six years ago. And anyone who is getting close to being in a leadership position has every right to feel more confident now than ever before about their ability to do the next job in their career. But the danger of AI is that it so silently and seamlessly fills in what we don’t know with whatever it thinks best to close the gap. It doesn’t know that we don’t know our prompt was underinformed. ChatGPT didn’t stop and say, ‘Robert, are you sure? This will look awful.’ My dance partner did (if only via smirk and live demonstration).
In the past, we hit our knowledge gap harder. Searching for routines online without AI would have probably revealed that I didn’t quite know what to look for. Ideally, I would have fatigued. In other words, those of us who don’t know enough would have self-eliminated trying to be experts. The search itself helped slow us down. Next gen leaders almost always have gaps in where they are and where they need to be (not to put too fine a point on it, but that is why they are next generation and not this one). They might even generally (hopefully) be self-aware enough to know the gap is communication, or technical skill, or relationship building. But the real work is in the space between kind of knowing you don’t yet know enough, and recognising it immediately, down to the last awkward pivot that this isn’t going to work.
The way next gen leaders will become ready probably hasn’t changed much in the era of AI. It’s really more of the same. You still need humans helping you reach the next level, because you don’t know what you don’t know. The issue is not AI itself. It’s operator error. You really can’t use AI to close gaps that you don’t know you have, and AI doesn’t know to stop and tell you. ChatGPT will not smirk. It will not pause and ask if you’re sure. It will hand you the next thing, confidently, in beautifully formatted prose, and the responsibility for catching what’s missing sits entirely with you. Which means the work hasn’t changed. Find the people who have more reps than you do. Show them what you’ve made. Let them smirk. That is how the gap closes. AI is useful for building on what you already know. The humans are still how you find out what you don’t.
Here is how to avoid operator error and actually leverage AI
The real use of AI professionally is probably the opposite of what you think. Not closing a big gap between what you don’t know and where you are — but instead, closing the small ones. When I ask AI about organisational psychology or leadership or corporate culture, I am leveraging decades of formal training and professional experience. I know in an instant if the AI result is off base. I can tell it to go back and do X and Y, and reconsider this and that. It doesn’t take much time because I know ‘instinctively’ what is wrong. It’s enormously helpful in finding that one quote by that one guy about the thing I already know. It can pull a good primary source that explains the psychology premise I already understand but don’t have a full text for. This, I believe, is how a next gen leader can improve themselves — pushing the edges of their strongest skills. For everything else, you risk tripping over your own two feet.
The competition is in two weeks. My dance partner has put together a wonderful routine for it.
