# AI Was Supposed To Take Your Job. Why Hasn't It? Maxinomics · 2026-07-01 Phil Andrews argues that AI tools like Claude and ChatGPT probably won't take your job, but someone using AI will if you don't use it. Opening with the 1967 debut of the ATM at Barclays, he shows bank tellers instead doubled over the next 50 years, peaking in 2007. Of every occupation title tracked by the U.S. Census over 60 years, only the elevator operator was fully eliminated by automation, while autopilot made airline pilots ten times more numerous. Citing MIT economist David Autor's research, he notes 60% of today's jobs didn't exist 50 years ago, because inventions that make work more valuable create new jobs rather than replacing workers. Using his 3D-printed Maslow's hierarchy pyramid, built with Claude, he explains competing against non-consumption, and footnotes that elevator operators retired because modern skyscrapers like the Burj Khalifa demand computer-managed elevators. ## Questions this episode answers ### How many jobs have been fully eliminated by automation in the last 60 years? Speaker 1 says exactly one job has been completely eliminated by automation over the past 60 years: the elevator operator, which last appeared in the 1960 U.S. Census. Meanwhile, 60% of jobs currently being worked didn't exist 50 years ago, per David Autor's MIT research. [8:53](https://maxinomics.podhood.com/2aab56e7-b0ef-4cad-8c8f-8e76425e4ced?t=533000) ### Why did ATMs increase rather than eliminate bank teller jobs? Speaker 1 explains that as the number of ATMs climbed, the number of bank tellers also climbed, doubling over the next 50 years until reaching a peak in 2007. The ATM let tellers expand what they did behind the counter rather than replacing them, surprising even bank executives. [0:40](https://maxinomics.podhood.com/2aab56e7-b0ef-4cad-8c8f-8e76425e4ced?t=40000) ### Why did automation make airline pilots more in demand instead of obsolete? Speaker 1 notes autopilot has existed since the 1920s, yet the number of pilots exploded because automation enabled more planes, bigger planes, longer routes and way more flights. Pilots remain as the final human who says the plane is landing safely, since the computer doesn't make the call when everything's on fire. [9:09](https://maxinomics.podhood.com/2aab56e7-b0ef-4cad-8c8f-8e76425e4ced?t=549000) ### What does 'competing against non-consumption' mean, and how did the host use AI with a 3D printer? Speaker 1 explains it's when a person can't find a solution that satisfies the job's requirements, so they do nothing instead. He couldn't justify a day designing a simple hook in CAD, but used Codex or Claude with a couple prompts and about 30 minutes, and nobody's job was taken because no pyramid designer existed. [14:11](https://maxinomics.podhood.com/2aab56e7-b0ef-4cad-8c8f-8e76425e4ced?t=851000) ## Key moments - **[0:00] The ATM story** - [0:00] The 1967 Barclays ATM debut: everyone assumed bank tellers were finished — instead teller jobs doubled for 50 years - [0:59] Phil Andrews frames the AI fear: ATMs, spreadsheets, and Claude all triggered the same mass-unemployment prediction - **[1:07] What people want** - [1:44] A Maslow pyramid Phil Andrews couldn't buy anywhere reveals the vast market of products nobody makes because demand is unproven - [3:44] Making one plastic pyramid costs $17,000 in molds — that's why the 'job iceberg' of unmade products stays hidden - **[3:45] What is a job?** - [3:58] What is a job exactly? Every job is the same: take an input, produce an output across a moat of complexity - [6:02] Lottery surveys show over half of workers would quit their current job with $10 million — people work to escape tasks, not to work - [7:05] Elevator operator vs airline pilot: the only job automation fully eliminated, and the one it made 10x more in demand - **[7:48] Elevators vs pilots** - [12:27] Why you don't own a 3D printer: CAD skills were the bottleneck until Claude designed Phil Andrews' desk hook in 30 minutes - **[12:28] The 3D printer** - [14:33] Competing against non-consumption: AI unlocks work people simply opted out of, so no job gets taken - **[14:35] 60% new jobs** - [15:22] MIT's David Autor found 60% of today's jobs didn't exist 50 years ago, and only one occupation was fully automated away - **[16:25] The bet** - [16:42] Spreadsheets expanded accountants, nail guns grew carpentry: Autor's data shows make-jobs-better inventions create new work - **[16:52] Footnotes** - [18:25] Phil Andrews' verdict on AI and jobs: 'No, but someone using AI will take your job if you do not use AI' - [19:20] Footnote: a 10% weekly AI productivity gain means 520% more output in a year — and AI still can't replace the hard creative parts ## Speakers - **Phil Andrews** (host) ## Topics Artificial Intelligence, Industry Economics ## Mentioned Amazon (company), Otis (company), ATM (product), ChatGPT (product), Claude (product), Codex (product), Gemini (product), Waymo (product) ## Transcript ### The ATM story **Phil Andrews** [0:01] On a sweltering June day, in front of a velvet curtain at Barclays Bank in North London, the man took a piece of paper out of his pocket and pressed it into the beckoning slot on the beige box. The crowd watched, murmuring amongst themselves with a confused awe as the box, seemingly happy with the paper it had been given, spit a crisp 10-pound bill into his waiting hand. As struck by this as the crowd, he slowly stepped away, inspecting the bill as though maybe it wasn't real. But the next person went, and the next, Barclays staff assuring the people not only was the money real, but that this was the way everyone would get money out of their account from now on. The murmurs turned into conversations, which turned into rumors, which turned into the nearly universal belief that the people whose job it was to hand cash to everyone through a window were definitely, without a question, about to be out of a job. Exactly like the conversation we're having today. This was the birth of the ATM, and it was obvious: why would we need bank tellers anymore? But as a surprise to just about everyone, including bank executives, everybody in the industry, as the number of ATMs climbed, the number of bank tellers climbed, doubling over the next 50 years until reaching a peak in 2007. ATMs, spreadsheets, photography, the printing press, Claude. Is AI going to take my job? **Phil Andrews** [1:04] This video is brought to you by Cometeer Coffee. More on them later. This is exactly what the scene called for: a perfect pyramid cut into five layers, each layer a different color, movable, able to be stacked one on top of another. ### What people want **Phil Andrews** [1:16] A visual representation of the five things every human needs to become the peak version of themselves. This layer,right above food, shelter, and water, is what this video is about: security. Your job. Solve security and you can move on to belonging, then to achievement, and should you be lucky enough, onto what they call self-actualization, the peak of human experience. And this pyramid is exactly the thing that no matter how hard I tried, I could not buy. That no one could buy because it simply did not exist in the seemingly infinite selection of things you can buy in the world. Not on Amazon, not Etsy, not even by negotiating directly from factories located in China. Because unless you know thousands of people want to buy exactly this type of pyramid, same colors, same size, it would be a pure gamble to go through the effort to make them. So it's a play game. You're sitting there thinking, "Huh, maybe I could make and sell them. That's kind of cool." How much do you think it would cost to make the first 1,000 of these? The pyramid is six inches tall by six inches wide. There are five pieces to it, and each piece is a different color. It's basic plastic. There is nothing special about the material. $2,000. $75,000. $5,000. **Phil Andrews** [2:17] $30,000. Okay. Allright. Well, really only want one. How about just one? $17,000. An aluminum mold engraved with an indent the shape of each layer to hold melted plastic needs to be made for each of the five layers. That's the cost of the molds, about $17,000. So to just break even on your first thousand, you'd have to sell them for $30 a piece. And even that doesn't take into account packaging time, everything that goes into getting this to a customer. So nobody makes it, and nobody bought it, and the market recorded nothing, a zero in a column that could have been a one. Me. Yet this pyramid is here. It's real, and it is the only one that exists. I did not know it at the time, but once I had it, this thing was the perfect representation of all the things we don't know people want that simply aren't created because there's too much uncertainty. It takes far too much time, money, both. Like the tip of an iceberg. Above the surface, what we know, what we can see, what's under it, we cannot see. Hidden by water, the true size of the iceberg is always much, much larger than what's visible to the eye. This pyramid, the tip of an iceberg, hints at the unknown yet infinite number of ways to arrange atoms and bits into things, experiences, that people find delightful, find safety, comfort, utility. Things that we take for granted. Like electricity. 250 years ago, nobody had any idea that spinning a bunch of copper around a magnet would produce lightning that could be harnessed and channeled into a little plug on the wall that lets you create winter in a box so you can always have fresh food. Things that might make people 250 years from now go, "They only live to like 87 back then. Can you believe that?" But amongst this optimism, the idea of infinite possibility represented by this pyramid, which I will explain how it came to be here in a minute, if we want to know if your job is going to be taken, we must ask the question, what exactly is a job? ### What is a job? **Phil Andrews** [3:58] Every job is exactly the same. Whether you're a nurse, writer, programmer, chef, whatever. You take something in and we output something different. Something goes in one side and something new comes out the other. **Phil Andrews** [4:09] The nurse takes in an unhealthy patient and outputs a healthy one. A writer takes in information and they output an article. The programmer takes in a process and outputs a more efficient one. Chefs take in vegetables, meat, out comes dinner. We push buttons, one button after another, to steer our ship through the moat of complexity that sits between what we have and what our customers want. **Phil Andrews** [4:30] Building PowerPoint decks and copying data from one spreadsheet into another is a true mind-numbing grind for 80 hours a week. And exactly what those that enter the consulting or investment banking world do for two to three years before being allowed to do any client-facing work. Learning how to make PowerPoint decks in very specific styles, flow, theright fonts, the pyramid principle, 10 to 20 rounds of revisions per slide. An employee is given the task they're given either because someone else doesn't want to do them or doesn't have time to do them. The question isn't, "Do they want to do this?" It's, "What needs to be done to turn the input into the output?" So that is how people see their job. Uber drivers watching Waymo pick up the customer they want. Programmers watching ChatGPT run laps around their handwritten code. Lawyers watching Gemini surface case law they didn't even know existed. Warehouse workers watching the figure robots sort for 200 straight hours. The pain, the agony, the horror. Or is it a horror? Do you even like your job? Uber drivers do not. Just four out of every 100 new Uber drivers are still driving after one year. Every tool ever built and sold to help coders was either directly or indirectly about having to write less code. Assembly to compile languages, keyboard shortcuts, autocomplete, open source packages. The lawyer didn't know the case law related to their case existed because they didn't want to spend the time searching and compiling thousands of pages of documents. To keep 100 roles staffed at a warehouse, Amazon has to hire 150 new workers every year. Not just newly opened warehouses, no, no. Every year, for every warehouse, 150 new workers will need to be hired to keep 100 positions full. So most people watching an AI tool do one of the things they have to do. That is not the job. That is the current required task that helps complete the job. These are quite often not tasks people actually like. Let me ask you a question. If you won $10 million in the lottery today, would you continue to work or would you stop working? Go ahead, take your time. Think about it for a second. Ready? It's an age-old question that's been asked in many surveys, so we know how people respond to this idea. More than half of people would, at minimum, not stay at their current job. Less than 25% say they wouldn't stop working for any amount of money. Money is this funny thing. Pieces of paper with faces printed on them so we can get something to eat without killing each other. It's not evil. It's not divine. It is simply the physical representation of infinite choice. What some would call optionality. A $20 bill will pass through the hands of a thousand people without being spent on exactly the same thing. Once it reaches your hands, you can choose to acquire an infinite number of things or experiences. Or, and this is the critical part, the ability to say no. Because clearly, from the decades of asking people, the number one thing people want money for so they can say no, "I don't want to spend my time like that." Which means this saying is wrong. Benjamin Franklin got it backwards, putting time first to emphasize that you should trade time to get money. That's what the whole paragraph you can find the saying in is about. When really, we know that humans work to obtain time. You want money to buy a house so you don't have to build it, to buy food so you don't have to grow it, to buy a car so you don't have to walk. That gap between the input you're given and the output you need to get is largely time spent doing things no one likes doing. The menus, clicking buttons, steering, tapping, messaging, sorting, waiting on hold, hammering a nail. Your job is to get the output from the input. The question most of us haveright now is really not the title of this video. It's this. And the answer to that? Probably. But your job? That depends on whether you're an elevator operator or an airline pilot. He used to hoist himself up on a platform, 30, maybe 40 feet above the ground. It was supposed to be a spectacle. ### Elevators vs pilots **Phil Andrews** [8:01] He actually designed the show with P.T. Barnum of Barnum & Bailey's Circus. He'd wave to the crowd and then an assistant would take a hatchet or axe and chop the thick rope cable that had been cranked to pull the platform up. The whole thing would drop a couple feet, the crowd would gasp, and then the brake that he invented would kick in and stop the platform from falling. That was how Otis marketed his invention, the elevator brake, the name you see in so many elevators. Without the brake, there would be no elevator. Humans had understood what elevators could do for a long time and somewhere in use for freight, but it was just too dangerous for passengers. If the cable broke, the whole thing would plummet. Very bad ending. Without the elevator, there would be no skyscrapers. Walking up and down 30 flights of stairs is not something you can sell to any prospective tenant. But these things, they needed to be manually operated at first. There was a lever that you would manually adjust to speed up and slow down, just like a gas pedal. You could go too fast, you could go too slow, you could miss the floor you wanted to stop at. Every elevator had a guy in it working the lever for almost 100 years. And it is the only job that has been completely eliminated due specifically to automation over the past 60 years. The U.S. Census keeps a list of all occupation titles. The last time an elevator operator was seen was in the 1960 Census. Right as the elevator operator was entering its prime. The airline pilot shows up. 1920. Hi, I'm a pilot. What do you do? I fly planes. Cool. How much of that do you actually do? Bro, almost none of it. Autopilot has been around since the 1920s. Fly-by-wire everywhere since the '80s. Auto-land, planes can literally land themselves in zero visibility. I spend most of the flight watching screens and drinking coffee. So you're basically obsolete,right? That's what they said. They automated this out of flying and then something weird happened. The number of pilots didn't go down. It exploded. More planes, bigger planes, longer routes, way more flights. Automation didn't kill the job. It made every pilot way more powerful. But what about when hits the fan? Bird strike at takeoff. Total hydraulic failure. Passenger has a heart attack at 35,000 feet. Some drunk hole tries to kick in the cockpit door. The computer doesn't go to jail. The computer doesn't make the call when everything's on fire. You do. I'm not there to turn knobs. I'm there to be the final human who says, "This plane is landing safely. Everyone's going home." Air travel went from rich people toy to hundreds of millions of normal humans flying every year. And they all want a highly trained human sitting up front. They automated 90% of my job and created 10 times more of it. Elevator operator, gone. Airline pilot, more in demand than ever. So tell me again, is Claude going to take your job? **Phil Andrews** [10:35] Hold that thought. I need a cup of coffee real quick. Little afternoon comment here is just what the doctor ordered. What are we going with this afternoon? Been going through the Brazil one. Always thought I was a dark roast guy, but the middle section. Medium roasts have been hitting the spot. I have just a few hours of light left. Want to make it count? This is the science behind making an incredible cup of coffee at home. A perfect system that does the hard part for you so you can make the best coffee at home in seconds without a machine or any skill. The former president of Nespresso invested after just one sip. Life's too short for bad coffee, even if you need that coffee on short notice. Hot, iced, espresso style. Heat a little water, pop out the puff, and melt it away with hot water. Or drizzle it overice and milk or water for iced coffee or iced lattes. Coffee this easy and delicious is possible because Commentier brews it at 10 times the strength of regular coffee. The moment it's done brewing, it's flash frozen from hot to negative 321 degrees Fahrenheit in less than two seconds. That seals all the flavor and aroma in so nothing gets lost. It is a perfect cup of coffee anytime you want it. It comes straight to your door on dryice in recyclable aluminum capsules, no plastic pods or machines. And this world mug collection is the perfect place to start. 16 different countries, 32 cups of coffee. You can do a little head-to-head match if you want. Award-winning small batch roasters that you probably know some of the names of. Intelligentsia, Onyx. I think you should start with the Brazil one, but you do you. You can get your world mug collection today at commentier.com/maxinomics and get $20 off your first order. Thank you to Commentier for the coffee and for sponsoring this video. Now let's get back to it. Given what we know about history and some very good data I'll present here in a second, it's probably not going to take your job. But change your job, that I would count on. Imagine if you walked into a new job today and you're like, "Actually, I'm good. I want a computer. I don't need one. It's not really my thing." Probably wouldn't fly,right? But fear not, there is a choice to be made and it is your choice. The cup can be viewed as either half empty or it can be half full. This pyramid, which has now appeared in two videos, represents the cup half full side. I 3D printed it. ### The 3D printer **Phil Andrews** [12:44] These things have come a long way since I last had one about 10 years ago, which was just a couple years after the 3D printing bubble had popped in a spectacular way. Companies that made these printers had sworn, "We're going to 3D print everything," people said. "Everyone will have one in their home." And of course that didn't happen. Why exactly didn't that happen? One of, if not the biggest reason, how many people do you know that can design in 3D space? Even something as simple as this hook. It slides on and off my desk. It holds my backpack. This is not that important. It's very useful to me. I stopped ripping my clothes on the nail it replaced. But I can't dedicate time to go back, remember how, and figure out how to design this kind of thing. Look at how simple this is. And I kid you not, it would have taken me a day to design this. Someone who has experience with CAD, all the CAD people out there are like, "A day?" Yes, for me. And at least a day for the pyramid. It's not the same as opening up Photoshop. There are dedicated programs you have to know, file formats, specific techniques for 3D printing. And the program I used to use was bought by another company. Now it's $1,000 a year. No, thank you. Could I learn a new one? Sure. But to get this geometry justright, perfectly to scale, smooth, that would have taken more time than I can justify. And it randomly hit me one day. I cannot remember how it crossed my mind. The file the design program spit out, it's just code. The 3D printer reads it like instructions and executes. It's not that different from a web browser. That seemed like something that Codex or Claude could help with. A couple prompts, maybe 30 minutes of time between design and setup, and the printer just ran and did its thing. This thing came out. I'm looking at it, I'm like, "This is awesome." Whose job got taken by this? Nobody's. There is no hierarchy of needs pyramid designer. I would have just figured something else out for these two videos. But this made both of them better. This is what the scene called for. The formal name for this concept is competing against non-consumption. When a person can't find a solution that actually satisfies the requirements of the job, they opt to do nothing instead. Let's come back to that census list of jobs for a second. This is such a rich set of information. Here's a sample of just some of the new jobs that have been created over the past 60 years. Airplane designer, tattooer, wind turbine technician, cybersecurity analyst, conference planner, mental health counselor, dog psychiatrist. ### 60% new jobs **Phil Andrews** [14:48] 60% of the jobs that are currently being worked didn't exist 50 years ago. 60%. Only one has been completely removed. The elevator operator. He has been called the academic voice of the American worker. David Autor is the one responsible for that 60% data point. His team at MIT went through 80 years of official U.S. Census records and spotted every brand new job title that suddenly appeared, like solar photovoltaic electrician or dog psychiatrist. But the brilliant part was then analyzing millions of patents to separate inventions that make existing jobs more valuable and powerful from ones that just replace workers. New jobs explode exactly where you think they do, where the makes-job-better inventions pour in. Let me give you a few examples. We've already seen that the ATM machine let bank tellers expand what they did behind the counter. In the bank world, stay there. Spreadsheets didn't remove the need for accountants. It wildly expanded them. Same thing with analysts. We invented the nail gun so carpenters could spend more time on interesting things rather than banging nails in an old day. Houses are much more elaborate because of that. If we could just type faster and print easier, we could write our articles faster and man, we would be done our work quicker. But the word processor didn't shrink the staff of the Wall Street Journal, and there are 10, maybe even 100 times more writers today because of it. Same thing with movies. To make a movie, the film from a camera, it used to be physically cut and spliced together, loaded back onto a big reels, like a big heavy process. Adobe didn't take away editing. It changed it. It made it far more powerful. Netflix became a global pastime. Augment or replace? That is the question every single one of us is wondering. Am I going to be augmented or am I going to be replaced? Having shown you everything I know about this topic, one of my experiences with it, I lean that it augments in a very powerful way. ### The bet **Phil Andrews** [16:25] And that statement, I lean, is critical. You should check my work. You should assume I'm biased. Same with David Autor. But the data and the history, they're very compelling. So if I were to conclude, if I were to bet put money on what will happen, the safest bet to make, the most likely answer to is AI going to take my job is no. But someone using AI will take your job if you do not use AI. Allright, I did not have a chance to record the footnotes before I left for vacation. And the crew is now taking a nap, so I have a quick break in the action to get these done. ### Footnotes **Phil Andrews** [16:59] And I've been thinking about them quite a bit. So here are the footnotes, the part that everybody keeps asking for and that everybody seems to enjoy quite a bit. Let's get to them. Here's the footnotes. Number one, part of me really wanted to soften the language I used for the close of this video. Someone using AI will take your job if you do not use AI. But I really do think that is accurate. It will, in short order, be the same as applying for a job that requires you to type, but you're just like, "I don't use computers." It just wouldn't make sense to hire that person. I also realize that I am a very, I guess the term is AI-forward person. This is why I feel comfortable saying what I do in this video. I'm an experimenter. I tinker and fiddle with all sorts of different things, and that allows me to see what is possible and also what is not possible. And what is not possible is a much larger swath than what is possible. So take solace in that if you do not like the idea of AI. It is that expectation of perfection coming out of AI that, in my view, holds people back from really digging into it. It's magic up to a very easy-to-find point. After that, it's about incremental productivity gains that are on a learning curve. It takes effort. Consider two people doing identical jobs. One gets a 10% productivity gain from AI every week. Just 10%. Even if those gains don't compound and it's just adding 10% every week, that's 520% more done in a year. If it compounds, that becomes a very big number very quickly. Number two, AI does not replace the hard parts. Thinking you will replace all your work with AI is a complete misjudge of what the stuff is capable of. It lets you move faster. Does anyone benefit from me searching through a 100-page PDF I read a couple months ago to find the relevant sections and numbers I need for research I'm doing? They definitely don't. I don't. No one else does. That is 20 minutes of my time that just disappears. It does not replace the hard parts. What you see here in these videos is exactly as difficult as it was before AI because it's still all human. The writing, flow, ideas, concepts, footage, the visuals. Christie Muldoon is a savant video editor. Seth Laupus and Xingpei Shen handmade all the incredible graphics you saw. And then there's a variety of people behind the scenes that fill the gaps necessary to bring this stuff to life. I assure you, not a single one of us has ever thought, "I bet we could do this with one less person because of AI." Number three, if you work on one of the top floors of the Burj Khalifa, it takes 10 to 15 minutes to get to the bottom of the building. That is why we don't build taller skyscrapers. We could physically construct them to be stable. There are challenges with that, but a major limiting factor is how fast you can get up and down from the top. 15 minutes at rush hour in a crowded elevator, switching cars on at least one floor, is about the limit. And a big part of the reason the elevator operator was retired. It wasn't just because, "Great, let's automate the elevator." The operator was not able to manage the complexity of how elevators needed to operate in modern skyscrapers, optimizing when to stop at which floor with 20 people in the car. And it would have been an expensive, probably inefficient way to operate the five-story elevator you find in basic hotels all over the world. The cost of that passed on to you, the traveler. There was no way to expand it. What additional responsibilities could you give the person stuck in the box that goes up and down all day? Thank you for watching. Like and subscribe, and I will see you when I get back from vacation. See you. --- This library is powered by PodHood (https://podhood.com), the podcast website platform.