top of page

The Leadership Skill AI Can't Replace: Knowing What to Solve

Nov 4
7 min read

Most of the leaders I work with are under pressure to get more out of AI, and some worry about what it means for their jobs. What I keep seeing is that the tools can't decide what problem is worth solving, and when teams aren't clear on that, they end up with a lot of work that looks good and doesn't help the business. That's also where I think leaders stay relevant, and it has less to do with technical skill than with knowing what you value and how you think.


I learned this firsthand. When my co-founder Josh and I started building Remix Career, an AI-powered career assessment, the technology made the building part fast. Neither of us is a software engineer, and we had a working product in about three months. The part that took real thought, and that AI couldn't do for us, was getting clear on what problem we were solving for someone in the middle of a career change. Whenever we were vague about that, the tools gave us something that technically worked and didn't help anyone, which was frustrating, because we'd spent the time and still didn't have something people could use. What finally made it useful was the part no tool could supply, which was the methodology I'd built over years of coaching, the questions I ask and the way I work with clients to get underneath what they say they want. AI could build the product around that thinking, but it couldn't come up with it.


I was talking with a senior leader at a large asset manager recently, and he described the same problem across his whole firm. They had rolled out Copilot and invested in teaching people how to prompt it, and within two or three months the newer tools had made a lot of that training less relevant. People were using it, he said, but for basic things, maybe 5 percent of what it can do. The skill that lasts is knowing what to ask, which depends on understanding the business well enough to see what needs solving, and as he put it, leaders have to give their people what to solve.


What if I'm training my own replacement?

There's a fear underneath a lot of this that I don't think leaders talk about enough. Many of the people I coach use AI every day, and some part of them knows that every prompt they write and every correction they make is teaching a system to do more of what they do. Even the people who built this industry feel it. In early 2025, Bill Gates told Jimmy Fallon on The Tonight Show that within the next decade humans won't be needed "for most things." He also said, "There will be some things we reserve for ourselves," and I think that's the question most leaders are really asking, which things those will be.


You ask AI for a first pass at something that used to take your team two days, it comes back in minutes and it's good, and for a second you feel relieved. Then there's a small drop in your stomach as you work out what that means for the person who used to do it, and then for you.


Jack Swift, a former Army Ranger and fintech CEO, talked about this fear when he was on my podcast, Midlife Remix. He sees AI taking over calculation and routine work as a kind of liberation, and he said it's important that people "step into that without fear, step into that position of co-creator."


I've run layoffs, and every time I had to let people go I felt terrible and couldn't sleep the night before, so I know the fear is real. When a company can do the same work with fewer people, it usually does, and some roles are going to go away.


The pattern I've noticed is that the work AI absorbs first is the work that can be fully described, the analysis, the first draft, the summary, the model. It's also getting good at spotting things, flagging which customers might churn or who on a team might leave. What it can't do is decide what that means for the business, make the call when the data points in two directions and own it afterward, or sit down with the person who's thinking about leaving and have the conversation that makes them stay.


Staying relevant mostly comes down to knowing what you value, what you see that other people miss, and what you're willing to put your name behind, and those are the things most of us have spent the least time getting to know in ourselves.


The research points the same way. When Microsoft Research and Carnegie Mellon surveyed 319 knowledge workers who use AI every week, the people who trusted their own judgment questioned and checked the output more, while the people who trusted the AI more did less thinking of their own.


What AI amplifies, and what it can't

Jack has thought about this as much as anyone I know. As CEO of TIFIN, he helped build an AI-native fintech company to a valuation near a billion dollars, and he has just written Level 7 Leadership: The Blueprint for the Human + Machine Era.


His book argues that AI amplifies whatever thinking is already there, so when a leader's thinking is clear, AI gives them real leverage, and when it isn't, AI mostly produces more noise, faster. Jack argues that the most important work for leaders now is internal, getting their head, heart and gut aligned before they act. He writes, "Leadership once meant having the answer. Now it means creating the conditions for new answers to appear." Those conditions are mostly about how a leadership team talks to each other, and I don't think you can build them in a training module.


When he was on my podcast, Midlife Remix, he said he doesn't see this as humans trying to beat machines, and put it simply: "it's not that we gain advantage over machines, it's human machine team."


He also makes the point that we aren't going to outwork machines, but we can out-trust our competitors. For anyone worried about being replaced, that matters, because trust is built between people over time, and a team can't get it from a tool.


Why sounding like everyone else stops working

I see this all the time. A lot of people spend real energy at work figuring out what their boss wants to hear and modeling themselves on whoever seems to be getting ahead, picking up the way they present, the language they use in meetings, even the opinions they think will go over well. That made a certain kind of sense before, but producing a reasonable version of what's expected is exactly what AI does well. What a team needs now are people who think for themselves and bring their own ideas and their own questions, and in my experience those are the people who get the most out of these tools. When I said this to the leader at the asset manager, he agreed, and said AI can amplify what you're really good at, as long as you stick to who you actually are.


When people say leaders need to be more human in the age of AI, it can sound soft, like a reminder to be nice. I mean something more practical. To direct AI well, you need to know what you want, what you value, what good looks like to you, and what it feels like in your body when something is off. For me it's a slight tightening in my chest when I read something that's technically right and isn't what I meant, and for a long time I ignored it because I couldn't explain it. If I can't tell what good looks like for me, I have no way of telling the machine.


The leaders and leadership teams I work with are asking their own versions of this. Which parts of our work should we hand to AI, and which should we protect? What am I actually bringing if the first draft of the strategy isn't mine anymore? How do we agree as a team on what good looks like before we ask the machine for it?


What I'm practicing

I don't have this figured out, but a few things have started to help. One is leaving some gaps in my day empty, a walk without a podcast, a few minutes with nothing to do. I don't mean checking out or going numb, I mean letting my mind settle enough to feel what's actually there, including frustration or worry or the sense that I've been saying yes to the wrong things. Those feelings tell me what matters to me, and that's what I need to know before I can tell AI, or anyone else, what I want.


That carries into how I use the tools. Before I open one, I take a few minutes to write down what I'm actually trying to solve and why it matters, even if it's messy, and sometimes those few minutes tell me the problem isn't what I thought, or that what I need is a conversation with a person. When something comes back, I read it slowly and notice whether that tightening in my chest shows up, and when something is off I try to say why, because the why is usually where my own thinking is.


I do something similar with teams. In brainstorming sessions, everyone spends five minutes writing down their own thoughts before anyone talks, then we work in small groups and bring it back to the whole group. Otherwise the loudest voices win, and the people who think before they speak, who often see what nobody else does, never get heard. The same is true with AI, because when people know what they think before they ask the machine, they bring something of their own to it, and the conversation with their colleagues afterward has more than one point of view in it.


I still catch myself asking AI for an answer before I've figured out the question, usually when I'm tired or rushing, trying to skip the part only I could do. That part is slower and messier, and it comes from my own experience, the clients I've sat across from, the mistakes I've made, the things I care about, and no tool has been through any of that. What do you see in your work that nobody else, and no AI, would see the same way, and how often do you actually trust it?

Comments


  • Apple Podcasts
  • Spotify
  • Instagram
  • Picsart_26-04-05_15-19-27-843
  • LinkedIn
  • YouTube

Get posts like this by email

Working through this yourself?

bottom of page