On my first client, I wrote hundreds of requirements by hand. I sat through hours of discovery sessions, listened closely to what people said they needed, and translated all of it into detailed documentation. It was demanding work and it took real focus to get right. At the time I saw it as the job. Looking back, it was also teaching me something bigger, which was what good actually looked like. I learned to hear when a stakeholder was describing a symptom rather than the real problem. I learned to spot the gap between what someone asked for and what they actually needed. In essence, I was learning how to be a consultant.

I have been thinking about that work lately, because an AI tool could now do a version of it in minutes. It could take the transcripts from those sessions and produce a clean first set of requirements faster than I could have written just a few of them. Which raises the question – if the work that taught me how to do this job is the same work AI now does, how does anyone learn to do it next?

A Different Seat Changes the View

I used to think about AI in terms of technology implementations. It was a tool that helped me get things done faster. It could draft the document, scaffold the early code, or summarise the research, and my job was to take that head start and turn it into something a client could actually use.

When I recently moved into a broader, AI-focused role, the picture changed. From a different vantage point, AI stopped looking like a tool to accelerate work and started looking like something reshaping how the work itself gets built, and who gets the chance to grow inside it. The closer you sit to strategy, the more you see AI as a question about people rather than a question about technology. That shift in view is what got me thinking about judgment, about leadership, and about where consulting actually goes from here.

What the Early Work Was Really For

In the traditional consulting pyramid, judgment was something you earned through experience. It never arrived all at once. It came from doing the unglamorous version of the work many times over. You drafted the thing, you got it wrong, someone more senior pointed out why, and slowly you built an instinct for when something was off before anyone had to tell you. The tedious work I did early on was the training ground, quietly teaching me what good looked like.

AI now does a great deal of that work, and the entry-level rung is being reshaped because of it. Across the industry, consulting firms are rethinking what junior roles even involve. They are restructuring around AI, redefining titles, and in some cases reducing headcount as the shape of the work changes. So the process that used to build judgment is breaking, and there is a second problem layered underneath it, which is the one I find more interesting.

You Cannot Practise a Job That Does Not Exist Yet

Even if we could perfectly recreate the old reps, we would be training people for work that is already changing in real-time. The roles coming next are roles we cannot fully describe, because the tools and the workflows that will define them are still being built. You cannot do a hundred reps of a job that has not been invented.

This is the part that makes the usual advice fall a little flat. We tell people to learn the fundamentals, and that still matters. But the fundamentals of consulting in three years may not look much like the fundamentals I learned. The skill is no longer just knowing the work. It is being able to keep learning the work while it refuses to sit still.

What Is Actually Scarce Now

For a long time, expertise meant what you knew and what you could do. Deep knowledge of an industry. The ability to build the model, write the code, structure the analysis. That was what separated strong knowledge workers from the rest. That definition does not hold anymore, because knowledge and capability are now remarkably accessible. A junior with the right tools has access to the same reasoning power and the same information that used to take years to accumulate. So expertise is being redefined. It is no longer about what you know or what you can produce. It is about something harder to acquire.

What is scarce is judgment and adaptability.

Judgment, because someone still has to know when the machine is quietly wrong. AI produces confident, well-structured output that is sometimes subtly off, and catching that requires an instinct you only get from having done the work yourself. You can direct an AI without ever having built the thing it is building. But then you lose the ability to sense when its answer does not look right, and the ability to sense that is increasing in value.

Adaptability, because the work will keep changing and the people who do well are the ones who can keep forming new instincts as it does. Judgment is the skill. Adaptability is what lets you keep building the skill when the landscape is always moving. Together, they are what expertise is starting to mean.

So What Do You Actually Do

There isn’t a straight answer, but here is what I have noticed in the people who are handling this well.

They use AI to clear the routine work out of the way, then spend the time it frees up making judgment calls they used to reach much later. It is not about doing less. It is about elevating yourself to the harder decisions sooner and making more of them, which means replacing the old reps with higher-weight, better ones. It also means surrounding yourself with people whose judgment you trust, the ones you can take a call to who will tell you plainly whether you have it right.

Take the requirements work I started with. Instead of spending weeks drafting it by hand, I can use AI to get a solid first version in a fraction of the time, which frees me to move to the questions that actually matter sooner. What would my manager challenge here? What is the next step I would normally not reach for another week, a client conversation or a quick prototype to test whether the idea even holds? I have written before about how AI is compressing delivery cycles and pushing teams toward faster iteration and rapid prototyping. What I have come to see more recently is that this reaches well beyond technology or the delivery model. Faster cycles, tighter loops, and better outputs, with your energy spent higher up the chain. You elevate yourself, and AI helps you get there.

They treat AI output as a draft to interrogate and iterate on, not a result to accept. Every time you catch where a model went wrong, that is a rep. Every time you improve upon an output by adding more of your own thoughts and context, that is a rep. Working with the AI itself is how you stay sharp.

They stay close to the edge of what the tools can do. The people adapting well are not waiting to engage. They treat constant change as the normal condition of the work rather than a disruption to wait out.

And if you lead people, the uncomfortable part is that you can no longer assume judgment will form on its own through exposure. The exposure is being automated away. So building judgment in your team becomes something you have to design deliberately, not something that happens because someone spent two years learning from you.

The Part That Is Actually Exciting

This is not a story about jobs disappearing. I think it is really a story about being open to change.

In the mid-1800s, more than half of the American workforce worked in agriculture. Today it is around two percent. The work did not vanish. It transformed, and most of what replaced it would have been impossible to describe to someone living back then. You could not have explained a financial analyst, a software engineer, or a management consultant to a farmer in 1860, because the economy that needed those roles had not been built yet.

I think we are standing somewhere similar. The work I learned on my first job is now largely done with the help of AI, and the work that comes next is work I cannot fully picture. I find that exciting. The people who learn to build judgment on purpose, and who treat change as the job rather than an interruption to it, are going to be the ones figuring out what the next job even is.