Why Replacing Humans with AI is Going Horribly Wrong

Added: 2026-06-01

[00:00:00]For years, the message was simple.

[00:00:02]Artificial intelligence was coming for human jobs. It was going to replace call center workers, office workers, junior employees, customer service agents, writers, coders, analysts, and maybe even doctors. Companies were told that AI would make them faster, leaner, and more profitable.

[00:00:22]Workers were told that if their job could be done on a computer, then eventually it could be done by a machine.

[00:00:28]But now the story is becoming much more complicated because the same AI revolution that was supposed to make human workers unnecessary is running into a very serious problem. It's expensive.

[00:00:39]It's messy. It's not always producing clear results. And now even Sam Altman, the CEO of OpenAI, appears to be changing his tone on the so-called AI jobs apocalypse.

[00:00:52]Times magazine reported, throughout his rise to becoming one of the most influential CEOs in artificial intelligence, OpenAI's Sam Altman made repeated bold assertions about the impact that the new technology would have on jobs.

[00:01:07]I don't think we're going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about. He said during a virtual interview as a Commonwealth Bank of Australia conference in Sydney on Tuesday.

[00:01:19]I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened, Altman said.

[00:01:28]I now think I understand more about why it hasn't, and I'm obviously grateful, but that is an area where my intuitions were just off.

[00:01:37]That is a major statement because this is not coming from someone outside the AI industry. This is not coming from a critic who never believed in the technology. This is coming from one of the most powerful figures in artificial intelligence.

[00:01:51]Altman is now saying that the impact on entry-level white-collar jobs has not happened as fast as he expected. And that matters because for years workers were told to prepare for mass disruption.

[00:02:02]Companies were told AI would slash labor costs. Investors were told this technology would unlock enormous productivity.

[00:02:10]But now we're starting to see the gap between the promise and the reality.

[00:02:14]Take a look at this clip where Sam Altman's past comments, just last year, about entire job categories disappearing are discussed, especially customer service and call center roles.

[00:02:26]>> Prediction one, certain job categories will vanish due to AI. That's what Sam Altman says. Once AI comes, some jobs will cease to exist. And he's even given some examples. He talks about jobs like customer service and hotlines, basically your call centers. They will go.

[00:02:41]>> Now, some areas, again, I think just like totally totally gone. I don't know if any of you have used one of these like AI customer support bots.

[00:02:51]But it's incredible.

[00:02:53]Couple years ago, you like call customer support, you like go through a phone tree, you talk to four different people, they do the thing wrong, you call back again, you wait through it. It's like hours of pain, ton of time on hold, and the thing that you want doesn't happen. Very frustrating experience.

[00:03:09]Now, you call one of these things, an AI answers. It's like a super smart, capable person. There's no phone tree, there's no transfers. It can do everything that any customer support agent at that company could do.

[00:03:20]>> That clip is important because it shows the contrast. On one side, you have the earlier warning, certain job categories could be totally gone.

[00:03:29]Customer support could be handled by AI.

[00:03:31]No phone trees, no waiting on hold, no transfers. A machine could supposedly do what a human support agent does, but faster and cheaper. But now the conversation is changed because companies are discovering that replacing humans is not just about whether AI can answer a question. It's about whether AI can solve the real problem, understand the customer, take responsibility, and do it at a cost that actually makes sense.

[00:03:59]And this is where the economics of AI become difficult. The early pitch was that AI would save money. Instead of hiring more workers, companies could use AI tools. Instead of paying salaries, benefits, training costs, and office expenses, they could pay for software.

[00:04:16]But that software is not free. It runs on expensive computing power. It uses tokens. It requires subscriptions. It needs cloud infrastructure. It needs engineers to manage it. It needs companies to track whether it's actually helping. And in some cases, the cost of using AI is starting to shock the very businesses that rushed into it.

[00:04:37]Take a look at this clip from CNN, where they discuss how some companies are now finding that AI costs can be higher than human labor costs.

[00:04:46]>> We can only hope, right?

[00:04:48]>> We were starting to see a prayer.

[00:04:49]>> Exactly, but that's what we're really starting to see, especially at the big tech companies. I spoke with a VP at Nvidia who first flagged this to me. He said, "Oh yeah, for months our costs for my team have been more for AI than humans." That was the first flag. And then we started to hear this coming out in droves. Uber's CTO said he already blew out his whole budget for 2026 just on AI-related costs. And obviously that means he's spending more on that than he's spending on human workers. And now I'm starting to hear especially from startup founders, they're bragging about their AI bills being high because it's kind of this sign of like, "Yeah, I'm really ahead in the AI race."

[00:05:24]>> It's a flex. I'm blowing so much cash on this.

[00:05:26]>> Exactly. Exactly. But the whole point of this was supposed to be that it cut down on costs, expanded profits, especially for public companies. But it's unclear if that's going to be tenable long term.

[00:05:36]>> That is the part many people missed. AI was supposed be the cheap replacement.

[00:05:41]But if a company's AI bill starts rising faster than its labor bill, then the whole argument begins to weaken. A human work has limits, but they also bring judgement, responsibility, and experience. An AI tool can run all night, but that also means it can burn through money all night. And when companies encourage employees to use AI constantly, especially for coding, automation, and agents, costs can spiral quickly.

[00:06:08]The danger is that businesses end up replacing workers with a system that still needs humans around it, still needs monitoring, and still costs a fortune to operate.

[00:06:18]Time's magazine reported, "Despite those numbers, there have been signs that some companies are struggling to find value in AI use."

[00:06:26]Uber's president and chief operating officer, Andrew Macdonald, also said in a recent podcast interview posted May 22nd, "It is becoming harder and harder to justify AI costs in the company."

[00:06:39]The company's chief technology officer, Praveen Nepali Naga, went viral in April for admitting Uber burned through its 2026 Claude code budget in just 4 months.

[00:06:51]That example is extremely important because Uber is not a small company experimenting on the sidelines. Uber is a major technology company.

[00:07:00]If even companies like that are finding it harder to justify AI costs, then the problem is real. This is not just about whether AI can do impressive things in a demo, it's about whether the numbers work at scale.

[00:07:12]A tool can look magical when one person uses it for one task, but when a large company rolls it out across thousands of workers, every query, every token, every automation, and every agent starts adding up. And suddenly, the miracle tool becomes another massive line item in the budget.

[00:07:31]This is why the return on investment question is becoming so important.

[00:07:35]Businesses are not charities. They do not spend billions on AI just because it's exciting. They spend because they expect a return.

[00:07:43]They expect fewer workers, faster output, better service, higher margins, and stronger growth. But, if AI creates more work, requires new systems, increases oversight, and produces unclear savings, then executives have a problem. They have to explain why they are spending so much money on a technology that may not yet be delivering the returns they promised.

[00:08:07]Take a look at this clip from CNN where they discuss the trillion-dollar question. How companies can actually prove AI is worth the money.

[00:08:15]>> What are you hearing then is the measure of this is worth it?

[00:08:20]>> Well, that is the trillion-dollar question, Kate. Like that >> The six trillion-dollar >> Hits it on the head because there's no clear ROI metric here. And I think part of it is that companies, again, they just want to signal that they are really ahead on this. But, I'm talking to some tech companies that are working on individual metrics for every single project to prove that the money they spent on the AI is leading to some sort of return. I talked to an executive at Salesforce who created an individual metric for every single project they were working on. That creates a lot more work. So, that also goes into the conversation about whether or not AI is letting us work less. Um but, those kind of metrics are going to be increasingly popular, I think, especially as some of these AI companies go public and there are bigger questions from investors about the ROI of all this investment.

[00:09:06]>> That clip gets to the heart of the issue.

[00:09:08]The problem is not that AI is useless.

[00:09:11]The problem is that companies rushed into AI before they had clear ways to measure its value.

[00:09:17]In some cases, they're now creating individual metrics for every project just to prove that technology is worth it. But, that creates more work. And that is the irony. AI was supposed to reduce work. Yet, in some companies, it's creating new layers of management, tracking, measurement, and justification.

[00:09:37]Instead of simply making everything easier, AI is forcing companies to ask basic questions they should have asked before spending the money. What are we using this for? How much does it cost?

[00:09:49]And what are we saving? And is it actually better than the human process it replaced? But the issue is not only financial, it's also human. One of the biggest mistakes companies made was assuming that work is just a list of tasks. Answer this email. Handle this complaint. Write this code. Diagnose this problem. Summarize this document.

[00:10:10]But human work is more than tasks. It's trust. It's context. It is accountability. It's knowing when something feels wrong. It's understanding tone, emotion, risk, and consequence. And even Sam Altman now seems to be acknowledging that the human part of work is much harder to replace than many people thought.

[00:10:33]Time's magazine reported, Altman went on to explain that the human part of employment could not be replaced by AI.

[00:10:40]And that people care about interacting with each other at work. "We really do care about our interactions with people," he added, which he said, "for better or worse, updated me to thinking that the job's picture is likely to be very different than we thought."

[00:10:56]That point is bigger than it sounds.

[00:10:58]People do not only want answers, they want reassurance. They want someone to listen. They want someone to take ownership. This is especially true when the issue is stressful, personal, expensive, or serious. A chatbot may handle a simple refund, but when a customer is angry, confused, scared, or facing a complex problem, the human part matters. There is something in a real human being that AI cannot fully copy.

[00:11:26]Call it instinct. Call it empathy. Call it judgment. Call it that human spark.

[00:11:32]Whatever it is, people can feel the difference. They know when they're speaking to a person who understands pressure, pain, frustration, and responsibility.

[00:11:41]And that is why many companies may discover that AI works best as a tool beside humans, not as a full replacement for humans.

[00:11:49]Take a look at this clip where Sam Altman talks about AI in healthcare, but then admits he would still want a human doctor involved.

[00:11:57]>> The OpenAI CEO says ChatGPT is better at diagnosing diseases than most doctors.

[00:12:03]But will Sam Altman trust it himself?

[00:12:06]Will he himself go to an AI bot instead of a doctor?

[00:12:09]The answer is no.

[00:12:11]>> ChatGPT today, by the way, most of the time can give you better It can It's like a better diagnostician than most doctors in the world.

[00:12:20]And I'll like like many people here probably put my symptoms in um and test results and you know, like there's all these stories on the internet of ChatGPT saved my life and you know, I had this rare disease and it found it and all these doctors didn't do it. And yet people still go to doctors.

[00:12:34]And I am not like maybe I'm a dinosaur here, but I really do not want to like entrust my medical fate to ChatGPT with no human doctor in the loop.

[00:12:42]>> That is the contradiction in one moment.

[00:12:45]AI may be powerful. It may be fast. It may even be better than some humans at certain narrow tasks. But when the decision really matters, people still want a human in the loop. They want a doctor, not just a chatbot. They want responsibility, not just a prediction.

[00:13:03]They want someone who can be questioned, challenged, trusted, and held accountable. And that is the problem with the idea of replacing humans completely.

[00:13:12]The more serious the work becomes, the more people care about who is behind the decision.

[00:13:18]This does not mean AI will have no impact on jobs. That would be naive.

[00:13:23]AI is already changing how people work.

[00:13:26]Some jobs will shrink. Some roles will be cut. Some companies will use AI as a reason to reduce head count.

[00:13:34]But the idea that companies can simply remove humans, plug in AI, and watch everything improve is starting to look far too simplistic.

[00:13:42]In reality, the companies that rushed too fast may damage customer service, create new costs, frustrate workers, confuse customers, and still fail to get the savings they expected. And that's is why replacing humans with AI is going horribly wrong in some places.

[00:13:59]Not because AI cannot do anything useful, it can.

[00:14:03]Not because AI will disappear, it will not. But because businesses believed the hype before they understood the limits.

[00:14:10]They treated people as costs to remove instead of understanding the value those people bring.

[00:14:16]They assumed that intelligence was the same as judgment. They assumed that automation was the same as trust.

[00:14:23]They assumed that cheaper software would automatically mean a better business.

[00:14:28]But now the bills are rising, the return is unclear, and even the leaders of the AI industry are admitting the jobs picture may be very different from what they once thought.

[00:14:38]The real lesson is simple. AI may change the world. But replacing humans is not as easy as the hype made it sound. Human beings are not just expensive machines.

[00:14:49]They are workers, problem solvers, communicators, decision makers, and sources of trust. And the companies that forget that may discover something painful.

[00:15:01]Removing humans could be much easier than replacing what humans actually do.

[00:15:05]But the danger is not only inside the workplace. The same AI hype that convinced companies they could replace humans is now helping to push the stock market into one of its most concentrated and fragile positions in modern history.

[00:15:19]Morningstar reported, after a bear market in 2022, the launch of ChatGPT sparked an AI frenzy that has taken concentration to new levels.

[00:15:30]The era of the Magnificent Seven and the hyperscalers saw Nvidia's NVDA market value hit an astounding $5 trillion in late 2025.

[00:15:41]According to CRSP data, the top 10 reached 37.7% of the US stock market value on October 31st, 2025, surpassing the previous month end peak of May 31st, 1932, when the top 10 stocks collectively represented 37.3% of market value.

[00:16:03]That is the central fact in this whole story. The top 10 stocks reached 37.7% of total US stock market value in October 2025.

[00:16:14]In other words, before this latest AI-driven surge, the record was set all the way back in May 1932, when the top 10 stocks made up 37.3% of the US market.

[00:16:26]That is why the Great Depression comparison is getting attention. But we have to be careful here.

[00:16:32]The point is not that today is exactly like 1932.

[00:16:36]The point is that the structure of the market is showing a similar warning sign.

[00:16:41]The market is becoming top-heavy.

[00:16:43]It's being pulled upward by a smaller and smaller group of companies.

[00:16:48]When that happens, the market can look strong on the surface, even if the strength underneath is much weaker.

[00:16:54]That is why concentration matters. A concentrated market can keep rising and make investors money, but it can also become fragile.

[00:17:03]If the biggest companies fall, the whole market can fall with them.

[00:17:07]The danger is that the stock market is no longer being lifted by broad strength across many companies.

[00:17:13]It's being held up by a small group of giants.

[00:17:16]If those companies disappoint investors or fail to meet huge expectations, the damage can spread quickly.

[00:17:22]This is part of why people compare today's market concentration to the Great Depression era. Back then, the market was also very top-heavy with a small number of major companies carrying a large share of the market.

[00:17:36]To be clear, concentration did not cause the Great Depression by itself. That crisis was caused by speculation, excessive borrowing, bank failures, collapsing demand, policy mistakes, deflation, and monetary contraction.

[00:17:51]But, concentration was one feature of that period, and that is why today's numbers are getting attention.

[00:17:58]Morningstar reported, "Looking at CRSP's 100 years of US stock market data, concentration levels used to be far higher.

[00:18:07]From the 1920s to the late 1960s, the weight of the top 10 stocks regularly exceeded 1/4 of overall market value.

[00:18:16]Then came a multi-decade broadening that spanned the bear market of the 1970s and the bull markets of the 1980s and 1990s.

[00:18:26]It wasn't until the late 1990s tech, media, and telecom bubble inflated the share prices of behemoths like Intel INTC, Cisco CSCO, and General Electric that the top 10 collectively crossed 20% share.

[00:18:42]The top-heavy market dynamic persisted for a couple of years beyond the popping of the bubble in March 2020. Similar to how the 1932 concentration peak came after the crash of 1929.

[00:18:54]A broadening trend started in earnest in 2003.

[00:18:59]This is important because it shows that concentration is not new.

[00:19:03]The market has been top-heavy before. In the 1920s, 1930s, 1940s, 1950s, and 1960s, the biggest companies often represented a very large share of the market. Then over time, the market broadened out. More sectors mattered.

[00:19:21]More companies mattered. The economy became less dependent on a narrow set of giants. But now that broadening has reversed. The biggest companies have taken over again.

[00:19:31]In the late 1990s, the internet bubble created a similar pattern.

[00:19:35]Investors crowded into technology, media, and telecom companies because they believed the internet would change everything. And in one sense, they were right. The internet did change the world. But the stock market still got carried away.

[00:19:50]Many companies became wildly overvalued.

[00:19:52]Some survived and became giants. Many others collapsed. That is the uncomfortable lesson for today.

[00:19:59]AI may be transformational.

[00:20:02]AI may change everything. That does not mean every company connected to AI is worth whatever investors are currently paying.

[00:20:10]Take a look at this next clip where economist Liam Halligan explains how obsessed equity markets have become with AI and tech stocks and how much of the market's value is now tied to the theme.

[00:20:22]>> Equity markets, those stocks and shares, are absolutely mesmerized with AI.

[00:20:30]And tech stocks and AI stocks now account for almost half of the S&P 500, which is the that the combined valuation of the 500 most important um listed publicly traded on stock markets companies in America. And even though they're relatively small in number, they now account for, you know, almost half of the total valuation.

[00:20:56]Up from a quarter less than 3 years ago.

[00:20:59]And and some companies, their valuations are absolutely massive.

[00:21:05]>> That clip helps explain why this matters. The stock market is not just excited about AI, it's mesmerized by it.

[00:21:13]Investors are treating AI as the next great industrial revolution, and maybe it is. But the issue is not whether AI is important, the issue is whether the market has already priced in too much good news.

[00:21:26]When a small number of AI-linked companies become responsible for a huge share of market gains, investors are no longer just buying the US economy.

[00:21:35]They're buying a very specific story.

[00:21:37]They're buying the story that AI spending will keep growing, data centers will keep expanding, chips will remain in huge demand, electricity will be available, profits will arrive, regulation will not slow things down, and the biggest companies will justify their valuations. That is a lot of assumptions. And when markets are priced for perfection, even a small disappointment can become a big problem.

[00:22:03]Reuters has reported, "The rise of the Magnificent Seven, US tech giants, has amplified equity market concentration.

[00:22:11]The top 10 US stocks currently account for 33% of the overall market value, according to Morgan Stanley analysts, and 37.5% of the MSCI USA index.

[00:22:23]It's even more extreme in some other tech-heavy markets.

[00:22:27]The top single stocks in South Korea and Taiwan, Samsung and TSMC, account for around 20% and 40% respectively of their benchmark indexes, Morgan Stanley figures show.

[00:22:40]In fact, these two companies alone account for a fifth of the entire MSCI Emerging Markets Index, which covers 24 countries."

[00:22:50]Reuters adds another important layer to this story. This is not only happening in America. Market concentration is becoming a global issue.

[00:22:58]In the US, the Magnificent Seven and other tech giants dominate. But in places like Taiwan and South Korea, individual companies like TSMC and Samsung carry huge weight in their local markets. That means global investors may think they are diversified across countries, but in reality, they may still be heavily exposed to a few giant technology companies.

[00:23:21]This is the key point. Diversification can become an illusion.

[00:23:26]You may own a global index fund. You may own an S&P 500 fund. You may think you own hundreds of thousands of companies.

[00:23:34]But if the biggest companies dominate the index, your returns will still depend heavily on what happens to a handful of names.

[00:23:42]Reuters reported, "The current period of concentration is mostly tied up to one theme, AI."

[00:23:48]This means the S&P 500 and Nasdaq and a growing number of indices in Asia have essentially become directional bets on the success of this nascent technology.

[00:23:59]In turn, if earnings and guidance from just a few tech giants fall short of expectations, the top-down drawdown could be as indiscriminate as the rally, potentially turning into a disorderly route, as battle-scarred investors know all too well.

[00:24:15]And given sky-high AI expectations, a market correction may not require AI to flop. The technology simply needs to be less revolutionary than expected.

[00:24:25]That line is one of the most important parts of this Reuters piece.

[00:24:29]A correction does not require AI to fail.

[00:24:32]That is the part investors need to understand.

[00:24:35]AI does not need to be fake.

[00:24:38]AI does not need to disappear.

[00:24:40]AI does not need to be useless. The market can still fall if AI simply turns out to be less profitable, slower to monetize, more expensive to build, or more difficult to regulate than investors currently expect.

[00:24:54]That is what happened in previous technology booms. The technology survived, but many valuations did not.

[00:25:00]Railways changed the world, but railway investors still suffered during speculative crashes.

[00:25:06]The internet changed the world, but the dot-com bubble still burst. AI may change the world, too, but that does not automatically mean today's stock prices are safe.