The Builders Are Alright

Almost every AI-and-jobs conversation I'm in lately ends in the same place: someone asks, mostly half-joking, whether they should be worried about their job going away. I usually tell them no, but then I have to explain what I actually mean by that, because the honest answer isn't "you're fine" and it isn't "we're all cooked." It depends on what your day actually looks like.

I'm pretty firmly anti-doomer about AI. I don't think it's coming for "knowledge work" as a single category. The category is too broad to be useful. A staff engineer shipping a new system, a PM running a launch, a designer cutting prototypes, and a person whose week is mostly forwarding documents and writing status updates are all "knowledge workers" by the same definition. AI does very different things to each of them.

The more interesting question isn't who gets displaced, it's what kind of work gets displaced. The short version of where I've landed: if you're a builder, AI is the most leveraged tool you've ever had. If most of your week is tangential to building, that's the part to pay attention to. Once you frame the question that way, it stops being scary and starts being a roadmap.

What I mean by builder

The way I think about it, a builder is someone who produces a shipped outcome and has authority over what gets produced. Both halves matter.

You can ship things without authority. Plenty of engineers spend their week closing tickets that someone else specced, designed, and prioritized. That's still valuable work, but it's mostly executing a plan that already exists. AI is very good at executing plans that already exist.

You can also have authority without shipping. Plenty of people spend their week in meetings deciding what other people should build, then handing off a doc and walking away. That's still valuable work, but most of the value lives in the handoff, and most of the handoff is information transfer. AI is also very good at transferring information.

The builders are the people doing both. They're deciding what should exist, and they're putting their hands on the thing until it does. PMs who actually ship the outcome instead of just owning the spec. Designers who sit in the codebase and make the change themselves. Engineers who care about whether the feature is right, not just whether the ticket is closed. Founders who do all of it because there's no one to hand it to.

This isn't really about job titles. It's about task types. Two people with the same role can be doing wildly different work day-to-day. Look at where your week actually goes. If most of it is writing, deciding, testing, and then shipping the thing, you're a builder. If most of it is moving information from one person to another, you're not, regardless of your title.

If most of your day is tangential, AI is your way out

If you read the last section and recognized yourself in the second category, the move isn't to panic. It's to look at your week and figure out which of those tangential tasks AI can do for you, and start doing them that way.

I wrote about this in How I'm Using AI to Replace Myself at Work. The title is a little tongue-in-cheek, but the underlying idea is real. A lot of what fills a knowledge worker's week is information transfer, summarization, and routine decision-making, and a properly set up AI workflow handles a meaningful chunk of it. Not perfectly, and not without supervision, but well enough that you can stop being the bottleneck for it.

The part most people miss is what to do with the time you get back. The wrong move is to use it to do more of the same kind of work, faster. The right move is to redirect it toward building. Spend it shipping something. Spend it sitting next to a builder and learning how they think. Spend it producing one concrete thing that didn't exist before you sat down. Even if your role doesn't formally include building, you can almost always find an outcome you have some authority over, and start treating it like the thing you ship.

The roles most exposed to AI are the ones where the entire job is the tangential work. If you can shift even a portion of your week toward building, you become much harder to replace, because the part that's left is the part AI struggles with: judgment, context, taste, and getting something out the door.

For builders, this is the most leveraged moment of your career

The flip side of all this is that if you already build, AI is rocket fuel.

I'll use myself as the example, because I'd rather get this wrong about my own work than someone else's. I've always been hands-on with the things I work on, and I take real pride in collaborating with strong engineers and designers. I still lean on them for the parts they're better at than me. The change isn't how involved I am, it's how directly my work shows up in the product and how quickly I can get a customer to react to it. Before, an idea I cared about had to travel through a long chain of people before it ever reached a customer. By the time feedback came back, I was several steps removed from the original instinct, and a lot of what made the idea sharp had been softened along the way. It was a slow game of telephone just to find out whether the idea was any good.

Now, when I have an idea, I can sit down and have a working prototype the same afternoon. I can put it in front of a real user the next day. I can have a credible spec, a working demo, and early customer reactions in a week, on something that previously would have taken a quarter and four people. My judgment about what should exist gets translated into a working version of that thing much more directly than it used to. My taste makes it into the product. My read on the customer makes it into the product. The latency between "I think this should exist" and "here's a version of it that exists" has collapsed.

That isn't free. The tradeoff is that I have fewer excuses. When the idea was sitting in a backlog waiting for capacity, I could tell myself the team would have built it eventually. Now, if it doesn't exist, it's because I didn't build it, or I haven't validated it well enough. The skill ceiling on being a builder has gone up, because the floor has too. Everyone has the same tools. What separates good builders from average ones is the same thing it always has been, which is judgment, taste, and the willingness to iterate when something isn't working. AI just makes those traits matter more, because they translate into output more directly than they used to.

The other thing this changes is how fast you find out you were wrong. Ideas don't just reach further. They get better, because the validation loop is shorter. I used to think an idea was good for months before anyone could test it. Now I can find out in a week that I was wrong about a feature, kill it, and move to the next thing. It's no longer "I have more ideas," but "I find out which of my ideas were bad, faster."

Both camps end up in the same place, just from different directions. Builders use AI to make their judgment count more, and people doing tangential work use it to free up the time they need to start building. The throughline is that the work AI rewards most is producing things, deciding what they should be, and getting them in front of someone who can tell you you're wrong.

If you're already a builder, this is the most leveraged your work has ever been, and the next few years are going to favor people who treat that as an invitation rather than a threat. If most of your week is tangential to building, this is the moment to find an outcome you can own, automate the parts that are getting in the way, and start shipping into it. The work won't always be the same shape it is today. The people who use this stretch to become builders, in whatever form their job allows, are going to be in a very different position than the people who wait to see how it shakes out.