Transcription
These companies all tried to replace humans. Tesla, Klarna, IBM, Duolingo and countless more. They were overconfident in replacing humans, but now, most are quietly backpedalling. Why are companies so obsessed with replacing humans? And, why haven’t all our jobs been taken by now? Strangest of all, one of the biggest disasters… was before the AI boom even began.
For our first example, we need look no further than Elon Musk, in 2017. This wasn’t AI like “Large Language Models”, but the entire experiment follows a very similar path. He had a grand vision, for a “machine that builds the machine”. A Tesla factory with as few human workers as possible. But, wait, isn’t that most car factories? Robots can move much more efficiently than humans, and almost every car factory already uses some robotics. They’ve also been like this way for years, if not decades, even by 2017. But this was different. Most automotive factories are a mix of robots and humans. But Elon Musk had become obsessed with efficiency and more importantly: automation. The colossal factory for the Model 3 would be the most robotics driven assembly line on the planet. Almost entirely robotics.
Also, the Model 3 was just launching, and expected to be Tesla’s big break. A 230-mile range, $35,000 EV, sedan, for middle class customers. The "bet the company" project. But, being more mass produced, means making a lot of cars. Musk wanted this car’s manufacturing, to be efficient, fast, and constant. The goal was 5000 Model 3s a week, by the end of 2017. Yet, that goalpost quickly began to move. The goal soon dropped to 2500. Then 2000. Production just wasn’t working. Cars couldn’t get out the door. Why? Tesla had created all kinds of bottlenecks. One, was the ‘systems integration subcontractor’, aka, the machine that makes batteries. This wasn’t working as intended, and Tesla had to rewrite the entire control software from scratch. But that was just the beginning. Some machines would break down 2 times a day, while the machine that lifted wheel well liners onto the production line would break down four times a week. The Robot inserting the front seats broke down as much as five times in a single day. Robotics in most automotive factories often go over a month between breakdowns. Apparently, “Tesla did not perform a lot of preventive maintenance on machines”, and with how many machines there were, it all compounded. The production line turned into a gigantic, complicated, bottleneck. And this was made all the worse by the cars themselves. Teslas aren’t known for being simple cars, and with how much new technology Tesla stuck in the Model 3, it pushed the robotics from bad, to disastrous.
[Musk]: “We put too much new technology into the Model 3 all at once. This -- this should have been staged,” he said. As 2018 rolled in, Tesla was far behind their targets. But, it gets even worse. In 2018, Tesla still hadn’t broken even in the past 9 years. Bankruptcy was a real, looming threat at this stage. When asked how close Tesla was to going under, Elon said [Elon Musk]: Closest we got was about a month. The Model 3 ramp was extreme stress & pain for a long time — from mid 2017 to mid 2019. Production & logistics hell. It was all a disaster. The machine that builds the machine, was going down, and it was dragging Tesla with it. Elon’s bet was on the verge of destroying his entire business. So what’d they do? They brought back humans. Elon scrapped pretty much the entire thing. Even with most of the robotics gone, they still had cars to make. Lots of them. So Tesla quickly built a temporary production line, called the “Sprung Project”. A gigantic, tent-like assembly line was raised in a parking lot. And it was jam packed with people, not robots. All assembling the Model 3, where the automation failed. And the unthinkable happened, or maybe very “thinkable”. Production accelerated. At long last, Tesla managed to get Model 3s out the door, at a good pace. They escaped bankruptcy. And the solution wasn't more automation. To be fair, Musk was very open about the failure. He tweeted [Musk]:“Excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated.”
This was before the AI boom, but it shows us something crucial about the shortcomings of automation and where exactly things can fall apart. Yet, very soon, thousands of companies would attempt the same idea, but not only replacing manual labour, but also, intellectual labour… As GPT, and soon other LLMs like Meta AI, Claude, Gemini, and Grok were rolling out, many people of all trades felt a sinking feeling in their stomach. Be it software, art, music, writing, customer support, or many more, they thought “will this replace my job?”, and unfortunately, their fears were soon realized. Business leaders were soon convinced that AI was finally here, and it was time for humans to go. Many CEOs began speaking of a new “AI first” future. The goal? More efficiency, more revenue, and most of all: Lower cost.
One of the most bold, was the buy now pay later giant Klarna. Its CEO, Sebastian, was amazed by GPT, and said he wanted Klarna to be OpenAI’s “favourite guinea pig”. I think you can see where this is going. In 2023, Klarna enacted a hiring freeze, and began laying off staff in droves. Staff dropped from 5,000 down to 3,800. Then, down to 2000. Staff were growing more concerned, but Sebastian had a very different tune. He bragged that AI chatbots were handling two thirds of customer service chats, within the first month, and that AI replaced 700 customer service agents. He also called the massive staff cuts “natural attrition”. In his words; [Sebastian Siemiatkowski]: "AI can already do all of the jobs that we as humans do. It's just a question of how we apply it and use it." But Klarna is just the tip of the iceberg. The Australian Telecom company Telstra cut 2800 workers in an AI overhaul. The Ecommerce giant Shopify had a similar tone. Leadership announced that all teams wanting to onboard more staff or resources, first had to show “why they cannot get what they want done using AI”. While these companies were making massive cuts, something else was happening in the background, which would cause problems… But for now, many CEOs were proclaiming this new revolution, including Micha Kaufman, CEO of Fiverr. In a leaked, company wide email, he told staff “Here’s the unpleasant truth: AI is coming for your jobs. Heck, it’s coming for my job too. This is a wake-up call”. This was after they began overhauling Fiverr to promote “AI services”, and launched an AI marketing campaign.
But these examples are nothing compared to the next: IBM. The old tech giant laid off 8000 staff. All replaced by AI. But, these were customer service roles. Many of these were HR roles, and IBM instead implemented “AskHR”, a digital, HR, Ai employee. It would reply to queries, perform documentation, and would approve or deny leave. Not only do we have AI reviewing our CVs, but now our time off too! But IBM insisted it was fine, because the overall role count was going up, saying [Arvind Krishna - IBM CEO]: “it gives you more investment to put into other areas”. These announcements and cuts outraged many, as the fears of thousands losing their jobs became a reality. But it wasn’t just big tech. Taco Bell replaced drive-through staff with an automated, AI voice and ordering system, and rolled it out at 500 locations in 2023. Though, the strangest out of all of these, was also the most unexpected. If you don’t know “DuoLingo”, it’s a gamified language learning app. But, this app is big. It’s a globally known brand, and not only that, a beloved brand. This bird is everywhere, and skits, animations and marketing stunts with Duolingo spawn tens of millions of views, and many marketing case studies. DuoLingo was doing well, really well. So, what happened next, came as a surprise. It would be “AI first”. Co-founder and CEO Luis von Ahn said [Luis von Ahn]: “[We] need to rethink much of how we work, minor tweaks to systems designed for humans won’t get us there”. Though he did sign his memo by proclaiming: “Duolingo will remain a company that cares deeply about its employees.” That didn’t extend to contract workers, though, who were being replaced by AI.
But Duolingo had no idea how drastic the consequences would be. After companies went full steam ahead with AI, the results that came were underwhelming. A study by researchers from MIT found that most new AI integrations don’t generate “rapid revenue acceleration”. In fact, only 5% do. And it’s important to note which 5%. [Aditya Challapally - Lead Author]: [Startups led by 19- or 20-year-olds, for example], “have seen revenues jump from zero to $20 million in a year. It’s because they pick one pain point, execute well, and partner smartly with companies who use their tools.” Some bigger companies pull it off, but 95% fail at integrating them. Leaders or managers treat them as “plug and play”. Layoff of staff and drop in an AI, but it doesn’t seem to work. ChatGPT might be able to adapt to how you work individually, but scaling that for systems and workflows of a giant enterprise? That takes a lot more work, and the report found that companies don’t seem to understand how to do that. So, dropping Claude or GPT into a large company can feel slow, or clumsy. They need data, pipelines, and training, and employees often don’t know how to make use of them.
But there’s more. A study by Orgvue of senior business leaders and executives, found that 55% of those who replaced employees with AI regretted it, and felt they made the wrong decision. Now, we’re seeing all of this, in the real world. Remember Klarna’s massive layoffs? And its CEO’s bold claims? Well, let’s take a look at them now. The company is now hiring again, including those 700 customer service roles. Why? You might be shocked to hear this, but Klarna saw a drop in quality after replacing staff with AI. Sebastian said: [Sebastian Siemiatkowski]: “As cost unfortunately seems to have been a too-predominant evaluation factor when organising this, what you end up having is lower quality”. Or in layman’s terms: “We prioritized cost over everything, and it didn’t work.” Surprise surprise, and they aren’t alone in their backpedalling. IBM leaders quickly realized the new “AskHR” bot wasn't as robust as they thought, and fell apart in the more complicated HR situations. IBM soon rehired many of the staff it laid off. Some of the outcomes though, were just odd. Another customer ordered a Mountain Dew, only for the AI to repeatedly ask him what drink he wanted with it. Others had unwanted items added, and were even charged twice. After all this, Taco Bell decided to scale back the AI system a bit.
While some are backpedalling due to underwhelming results, others are because of backlash. Duolingo’s “AI First” policy didn’t sit well with users, who saw a decrease in quality, and let the app know. “I cancelled my subscription after that news. There's at least some content in the English to Spanish course now that's AI generated, and it sucks. “We don’t want AI, we want real people doing good work. Goodbye, Duo.” “A 650+ day streak never felt so meaningless once I saw the news.” The CEO quickly rushed into damage control, saying he “did not give enough context”, and that Duo wouldn’t lay off any full-time staff. But, Duo still appears to be relying on AI regardless. So the question is, how and why, did all this happen? Why didn’t AI replace humans? All of the CEOs and leaders seem to be chasing the “illusion” of the perfect company. And I can’t entirely blame them. It’s good to try and make your business as efficient and productive as possible, and business leaders often talk about “automation” and “systems.” But there’s something deeper than all of this. Deeper than just “AI bad, humans good”. Which is fragility. Automation promises efficiency, but quite often, it creates fragility. All of these AI integrations and automations, that seem like they can replace workers, are actually quite fragile. And something that AI definitely can’t replace is subscribers. We’re trying to cross 1 million by the end of this year, and we’re over 80% there. If you hit subscribe right now, we might just be able to pull it off. A couple of robots malfunctioning bottlenecks the entire tesla factory, or one customer asking for 18,000 water cups. The weakest link can bring down the entire system, or bottleneck the production line. Humans add resilience to companies. In a way, we’re shock absorbers. When something goes wrong, we can adapt, change, press in, and ease off. When an HR situation requires nuance, we can provide it. We can jump from one job to another. And we don’t crash when someone asks for a lot of water cups. Humans are underrated.
It’s hard to say if AI will be able to do any and all types of jobs. But there is some hope. The Orgvue study also found that “51% of leaders believe reskilling is strategically important in preparing their workforce for AI and 41% say they have increased their L&D budgets to ensure employees have the right training.” Giving employees better tools and training, instead of laying them off, and dropping AI as “plug and play” is likely the better way forward. The startup that has felt the brunt of this more than anyone else is Fiverr, whose stock has crashed over 90%. Check out this video to learn more.