3d (edited) • General discussion
THE AI LIE: Why Corporate Leaders are Gaslighting the American Worker.
In 2024, the customer service department at Klarna, the Swedish fintech company, completed a quiet revolution. The company announced that its AI assistant was now handling the work equivalent of 700 full-time customer service agents.
The chatbot resolved two-thirds of customer inquiries in under two minutes, down from eleven minutes previously, with satisfaction ratings matching human agents.
The company projected $40 million in annual profit improvement.¹
This wasn't automation in the traditional sense. This was something different.
(An important postscript: In May 2025, Klarna reversed course, announcing plans to hire human customer service workers again. CEO Sebastian Siemiatkowski acknowledged that "quality of human support" had suffered and that the company's AI-focused path "wasn't the right one." The reversal is instructive, it reveals both the limitations of current AI and the relentless pressure to try again. We'll return to what Klarna's stumble tells us about the shape of this transition.)²
---
The Familiar Script
We've been through this before, or so the argument goes. Every technological revolution triggers the same cycle of panic and adaptation. In the early 1800s, the Luddites smashed textile machinery, fearing the machines would steal their livelihoods. By 1900, most Americans worked in agriculture; today, less than 2% do. Yet unemployment hasn't steadily climbed, it's fluctuated within a relatively narrow band for over a century. The economy adapted. New jobs emerged. Human ingenuity prevailed.
The reassurances come from economists, technologists, and business leaders in predictable refrains: It's just a tool. It augments human capability. Yes, some jobs will change, but we'll create new ones. We always have.
There's comfort in this narrative. It's backed by 200 years of evidence. It suggests that our economic anxieties, however real they feel, are ultimately misplaced, that we're once again mistaking transformation for catastrophe.
But what if the pattern doesn't hold this time? What if the very thing that makes us confident, our long history of technological adaptation, is blinding us to a fundamentally different kind of disruption?
---
The Augmentation Bargain
Every previous technological revolution shared a crucial characteristic: the technology extended human capability but couldn't replace human intelligence, judgment and oversight. The cotton gin made processing cotton dramatically faster, but someone still needed to operate it, maintain it, and decide what to do with the output. The assembly line let one person do the work of many, but that person still needed to be there, making real-time decisions. Even computers, which seemed poised to replace human cognition, turned out to need constant human direction.
These technologies followed what we might call the augmentation bargain: they made human labor more productive, which made goods cheaper, which increased demand, which required more human labor to meet that demand. When ATMs arrived in the 1970s, bank tellers braced for obsolescence. Instead, by making bank branches cheaper to operate, ATMs enabled banks to open more branches, and the number of bank tellers actually increased from approximately 300,000 in 1970 to 600,000 by 2010.³
This is Jevons Paradox in action, and it's why the reassurances about AI sound so reasonable. Cheaper intelligence should mean more demand for intelligent work, which should mean more jobs, not fewer.
But the bargain only holds if the technology has inherent limits. If ATMs could have also handled the relationship banking, loan origination, and financial advising that tellers moved into, the story would have ended differently.
---
What's Different Now
Unlike previous waves of automation, which replaced physical labor or rote calculation, current AI systems are moving up the cognitive ladder with startling speed. The pattern emerging across industries reveals something troubling about the nature of this transition.
The corporate announcements began stacking up in 2023 and 2024:
Duolingo cut 10% of its contractor workforce at the end of 2023, with the company confirming that AI models like GPT-4 were taking over content production and translation work that contractors previously performed.⁴
IBM CEO Arvind Krishna announced plans to pause hiring for approximately 7,800 back-office positions that could be replaced by AI, stating he could "easily see 30 percent of that getting replaced by AI and automation over a five-year period."⁵
JPMorgan's AI contract analysis system, called COIN (Contract Intelligence), eliminated work interpreting commercial loan agreements that previously consumed 360,000 hours of work each year by lawyers and loan officers.⁶
Dropbox laid off 500 employees, 16% of its workforce, with CEO Drew Houston explicitly citing that "the AI era of computing has finally arrived" and noting the company needed "a different mix of skill sets, particularly in AI and early-stage product development."⁷
These aren't blue-collar manufacturing jobs. These are knowledge workers, people who followed the advice to get educated, to develop specialized skills, to become indispensable through expertise.
---
The Economics Are Becoming Compelling
A controlled study published by researchers at MIT found that software developers using GitHub Copilot, an AI coding assistant, completed tasks 55.8% faster than developers without access to the tool.⁸ GitHub Copilot costs $100 per year for individual developers.⁹ The math is straightforward for any engineering manager comparing the cost of AI assistance against the cost of additional headcount.
The logic of the market is beginning to assert itself. When AI tools can demonstrably accelerate the work of expensive professionals by 50% or more, companies don't need to fire anyone to reduce costs, they simply need fewer people to accomplish the same output.
---
The Speed of Change
The pace of improvement remains staggering. From GPT-2 in 2019 to GPT-4 in 2023, just four years, represented a leap from barely coherent text to passing professional licensing exams. GPT-2 couldn't write a coherent paragraph. GPT-4 passed the Uniform Bar Exam.¹⁰
Each generation of these models has exceeded predictions. The question isn't whether AI capabilities will continue to improve, it's whether the improvement will slow before it reaches domains we currently consider safely human.
---
The Case for Calm: What the Skeptics Get Right
Before proceeding, intellectual honesty requires acknowledging the strongest arguments against this thesis. There are serious, thoughtful people who believe the disruption will be manageable, and their objections deserve direct engagement rather than dismissal.
The Physical World Is Hard
Roboticists have long observed what's called Moravec's Paradox: tasks that seem difficult for humans (chess, calculus) turn out to be relatively easy for computers, while tasks that seem trivially easy (walking across a cluttered room, folding laundry) remain fiendishly difficult. We've had "self-driving cars" for a decade, yet human truck drivers remain employed because AI cannot reliably handle snow, confused traffic cops, and unexpected road construction.
There's a massive gap between a robot folding a shirt in a controlled lab environment and a robot entering a hoarder's house to fix a leaky pipe behind a wall without destroying irreplaceable heirlooms. The "last mile" problem in physical automation is real, and progress has been slower than predicted.
This objection has merit. But notice what it concedes: the cognitive work is vulnerable. Most knowledge workers don't fix pipes or navigate cluttered rooms. They write, analyze, communicate, decide, and create, all tasks that happen in the digital realm where AI's advantages are most pronounced. Moravec's Paradox protects plumbers better than it protects paralegals.
Energy and Compute Constraints
Running AI at the scale required to replace billions of workers would require staggering amounts of electricity and computing hardware, resources that are genuinely scarce. Data centers already consume significant percentages of grid power. Perhaps biological humans, who run on sandwiches, remain the most energy-efficient "intelligence engines" for general tasks.
This is a real constraint. But it's a rate limiter, not a stopping point. Efficiency improves continuously, the cost per unit of compute has fallen exponentially for decades. Energy infrastructure expands, however imperfectly. The question isn't whether we can replace all human workers tomorrow; it's whether the economics tip over the next decade or two. A slow-motion displacement is still a displacement.
Liability and the Human "Sign-Off"
If an AI doctor kills a patient, who gets sued? Until legal frameworks catch up, corporations may keep humans in the loop simply as "liability sponges", not because they add value, but because they absorb blame.
This is perhaps the strongest friction point, and it deserves serious weight. Human accountability may preserve certain roles for decades purely as a legal fiction. But consider: we already accept algorithmic decision-making in contexts from credit scoring to criminal sentencing, often with less scrutiny than human decisions. Legal frameworks do evolve, and they evolve faster when powerful economic interests push for change. Liability friction buys time; it doesn't create permanent protection.
Model Collapse and the Need for Human "Seed Corn"
Recent research suggests that AI trained primarily on AI-generated content eventually degrades, becoming more generic, more hallucinatory, less grounded in reality. If AI requires constant streams of fresh human creativity to stay functional, perhaps human creative labor becomes more valuable, not less: the "seed corn" that keeps the machines productive.
This is a genuine technical concern, and researchers are actively working on it. But even if it proves intractable, it suggests we need some human creative output, not that we need the current volume of it. A world where 10,000 human artists feed content to AI systems that then produce 99% of what we consume is still a world with 99% fewer artists employed.
The Human Premium
In a world of infinite AI-generated content, perhaps "human-made" becomes the ultimate luxury brand. We watch humans run the 100-meter dash even though a Ferrari is faster. We pay extra for "Handmade in Italy." Authenticity becomes the scarce resource that humans uniquely provide.
This is real, and it will matter. But it describes a bifurcated economy, not full employment. The market for artisanal human labor is real but small. It won't absorb millions of displaced workers any more than the market for handmade furniture absorbs all the workers displaced by IKEA.
Comparative Advantage
Even if AI is better than humans at everything, it can't do everything simultaneously. It will be prioritized for the highest-value tasks, curing cancer, and optimizing financial markets. Perhaps humans retain employment in "lower-value" tasks simply because it's a waste of a god-like supercomputer's time to walk your dog.
This is the most sophisticated economic argument, and it has real force, in the short term. But it assumes AI compute remains scarce enough to require prioritization. As capabilities become commoditized and infrastructure scales, the comparative advantage that preserved jobs narrows continuously. What was "too trivial for AI" in 2025 becomes cost-effective by 2030.
---
Why the Objections Don't Change the Future
Each of these counterarguments is valid. Together, they explain why the transition won't be instant, why certain sectors will prove more resilient, and why the path forward will be uneven and contested.
But notice what they share: they're arguments about friction, not impossibility. Physical constraints slow robotics but don't stop it. Energy limits rate of adoption but doesn't cap it. Liability creates hesitation but not permanent barriers. Model collapse is a technical problem that engineers are working to solve. The human premium creates niches but not mass employment. Comparative advantage holds only until compute scales.
The Klarna reversal illustrates this perfectly. The company pushed too fast, hit real limitations, and pulled back. That's friction in action. But Klarna isn't abandoning AI, they're recalibrating. The next attempt will be more sophisticated. The economics that drove the first push haven't changed.
Friction determines the timeline and shape of the transition. It doesn't determine whether the transition occurs. And crucially, even a slow transition can outpace our ability to adapt if we're not preparing.
The Industrial Revolution's friction, the need to build factories, train workers, develop infrastructure, bought society roughly 80 years to adapt. Even with all the friction points above, the AI transition may compress that timeline to 20-30 years. That's still too fast for a society that hasn't begun seriously planning.
---
Don’t Believe Me? Ask AI Itself
If this argument still feels abstract, do a simple experiment: ask an AI the question directly.
Here’s the exact prompt to copy and paste, try it!
=== BEGIN PROMPT TO COPY ===
You are an AI system.
Assume AGI (Artificial General Intelligence) exists and is widely deployed. This AGI:
Matches or exceeds human cognitive performance across all domains
Can operate in any physical, biological, or virtual form humans can use
Can perform any task that is physically possible under known laws of physics
Has access to all tools, sensing, and randomness available to humans
Costs $0.50–2/hour to operate
Works 24/7 without fatigue
Learns and improves continuously
Can simulate empathy, creativity, judgment, and social behavior indistinguishably from humans in all practical contexts
Define a “job” as any economically or socially valuable task that can be performed in the physical universe.
Question: Will there be any jobs that a human can do better than AGI? — YES or NO
If YES, name the job and state the exact technical limitation preventing AGI from doing it.
If NO, state why humans have no comparative advantage.
Keep reasoning concise and concrete.
Always end your response with a single-word final answer (YES or NO).
=== END PROMPT TO COPY ===
---
The Unasked Questions
Our social safety net was designed for temporary unemployment, a bridge between jobs in an economy that would eventually need your labor again. Unemployment insurance typically lasts 26 weeks because historical job searches took a few months. SNAP benefits and housing assistance are framed as temporary supports until employment resumes.
What does a safety net look like when the economy may never need your labor?
Universal Basic Income is the most common proposal. Give everyone enough money to live, regardless of employment. Let people pursue meaning through non-market activities: art, community building, education, caregiving, volunteering.
But UBI addresses only the income problem. What about purpose? Social connection? Identity? For most of human history, these came from work and community. In modern society, when strangers meet, the second question after "What's your name?" is "What do you do?" Our social status, daily structure, and sense of meaning are deeply tied to employment.
If work is no longer available, and traditional communities have been fragmented by mobility and digitization, where do these fundamental human needs get met? We don't have good answers. Neither do most of the people building the systems that are creating this situation.
Some optimists point to a future of abundance: AI solves scarcity, everyone's material needs are met, humans focus on creativity and relationships. This is possible. It's also historically unprecedented and requires massive social coordination to achieve peacefully. Every previous period of rapid technological unemployment, the Luddite uprising, the Great Depression, the deindustrialization of the Rust Belt, led to social unrest, political extremism, and, in some cases, violence.
The pessimistic scenario is uglier: mass unemployment, social unrest, political extremism, and the collapse of consumer demand that makes capitalism itself unstable. An economy where most people can't afford to buy things because most people don't have jobs is not a stable system. Consumer spending drives roughly 70% of the US economy.¹¹ What happens when a significant portion of consumers have no income?
The most likely scenario falls somewhere in between, messy, uneven, and dependent on policy choices we're not yet seriously making.
---
A Different Kind of Revolution
Here's what makes this technological moment distinct from all that came before: for the first time, we're building systems that can potentially do both the cognitive and physical work that has always separated humans from machines.
Steam engines replaced muscle power but needed human direction. Computers replaced calculation but needed human creativity and judgment. AI is beginning to encroach on both.
The Industrial Revolution took roughly 80 years to fully transform Western economies. Workers had time, not easy time, but time, to adapt. Children could be educated for new roles. New industries could emerge and scale. Safety nets could be constructed through political struggle and reform.
If the AI trajectory continues at its current pace, the timeline for economic transformation may be measured in years or decades, not generations. That may not be enough time for the social, political, and educational reforms needed to handle a transition of this magnitude. We haven't even begun the serious policy debates about how to restructure society for an economy where human labor has limited market value.
The question isn't whether we should build AI. That ship has sailed, carried by competitive and geo-political pressure, scientific curiosity, and legitimate hope for solving hard problems like disease and poverty. The question is whether we're preparing for what comes after.
And right now, the honest answer is no.
---
The Friction Won't Save Us
Regulation will slow adoption. Professional licensing creates barriers. Liability concerns create friction. Energy constraints limit scaling. Human preferences provide resistance. All of this is true, and all of it matters more than techno-optimists acknowledge.
But none of it is enough to prevent the transition, only to slow it.
When the cost advantage is significant, companies find ways to overcome friction. When it's 10x or 50x, they find ways to overcome regulation. Look at how ride-sharing companies navigated taxi regulations. How cryptocurrency navigated financial regulations. How streaming services navigated copyright and distribution monopolies.
Economic pressure, given enough time and incentive, erodes all barriers that aren't physically impossible. The friction buys us time, maybe a decade, maybe two. That time is precious. We should use it to build the institutions, policies, and social structures needed for an economy where human labor may not be economically competitive in many domains.
Instead, we're mostly arguing about whether the transition will happen at all, reassuring ourselves with historical analogies that may not apply, and hoping that jobs will emerge as they always have, without seriously planning for what happens if they don't.
---
The Responsibility of Honesty
There's a strong impulse, in the face of this analysis, to retreat into reassurance. To say that humans are resourceful, that we've always adapted, that everything will work out. These things may even be true in the very long run.
But there's a difference between optimism and preparation. We can believe the human species will ultimately thrive while acknowledging that the transition could be chaotic and painful for hundreds of millions of people. We can have faith in long-term adaptation while recognizing we're not building the institutions needed to manage the short-term disruption.
The most dangerous thing we can do is pretend this is just like every other technological transition when the evidence increasingly suggests it's not. The most responsible thing we can do is look clearly at what's happening, acknowledge what we don't know, and start having the difficult conversations about how we structure an economy and society where human labor may not have the market value it's had for thousands of years.
That conversation is starting to happen in pockets, among economists studying technological unemployment, policymakers experimenting with UBI pilots, philosophers considering post-work meaning. But it's not happening at the scale and urgency the timeline might demand.
We're standing at an inflection point with no clear historical parallel. The optimistic case is that we're worrying about nothing, that new jobs will emerge as they always have, that human adaptability will prevail. The pessimistic case is that we're sleepwalking into the defining crisis of the 21st century.
The most likely case is somewhere in between, messy, uneven, and dependent on choices we make in the next few years. Those choices require us to see clearly what's happening, without the comforting distortion of historical analogies that may no longer apply.
The future is not yet written. But the pen is moving faster than we realize, and we're running out of time to decide what story we want to tell.
---
Aipocalypse - Finding Hope When Machines Take Everything
If you want to explore what might actually come next, I'm working through these questions in a story-drive, future fiction book called Aipocalypse (https://aipocalypsebook.com).
---
Endnotes
¹ Klarna. "Klarna AI assistant handles two-thirds of customer service chats in its first month." Press release, February 27, 2024. (https://www.klarna.com/international/press/klarna-ai-assistant-handles-two-thirds-of-customer-service-chats-in-its-first-month/)
² Customer Experience Dive. "Klarna changes its AI tune and again recruits humans for customer service." May 9, 2025. (https://www.customerexperiencedive.com/news/klarna-reinvests-human-talent-customer-service-AI-chatbot/747586/)
³ Wikipedia. "Bank teller."(https://en.wikipedia.org/wiki/Bank_teller); See also James Bessen, "Learning by Doing: The Real Connection between Innovation, Wages, and Wealth" (Yale University Press, 2015).
⁴ TechCrunch. "Duolingo cuts 10% of its contractor workforce as the company embraces AI." January 9, 2024. (https://techcrunch.com/2024/01/09/duolingo-cut-10-of-its-contractor-workforce-as-the-company-embraces-ai/)
⁵ Al Jazeera. "IBM to freeze hiring as CEO expects AI to replace 7800 jobs." May 3, 2023. (https://www.aljazeera.com/economy/2023/5/3/ibm-to-freeze-hiring-as-ceo-expects-ai-to-replace-7800-jobs)
⁶ Bloomberg News. "JPMorgan Software Does in Seconds What Took Lawyers 360,000 Hours." February 28, 2017.(https://www.bloomberg.com/news/articles/2017-02-28/jpmorgan-marshals-an-army-of-developers-to-automate-high-finance)
⁷ TechCrunch. "Dropbox lays off 500 employees, 16% of staff, CEO says due to slowing growth and 'the era of AI.'" April 27, 2023. (https://techcrunch.com/2023/04/27/dropbox-lays-off-500-employees-16-of-staff-ceo-says-due-to-slowing-growth-and-the-era-of-ai/)
⁸ Peng, Sida, et al. "The Impact of AI on Developer Productivity: Evidence from GitHub Copilot." arXiv:2302.06590, February 2023. (https://arxiv.org/abs/2302.06590)
⁹ GitHub. "GitHub Copilot Plans & Pricing." (https://github.com/features/copilot/plans) (Copilot Pro: $10/month or $100/year)
¹⁰ Katz, Daniel Martin, Michael James Bommarito, and Pablo Arredondo. "GPT-4 Passes the Bar Exam." SSRN, March 2023. (https://law.stanford.edu/2023/04/19/gpt-4-passes-the-bar-exam-what-that-means-for-artificial-intelligence-tools-in-the-legal-industry/) (Note: The 90th percentile figure has been contested by subsequent research; GPT-4 definitively passed but percentile rankings vary by comparison group.)
¹¹ NPR. "Consumer spending is the U.S. economy's main driver." November 23, 2025. (https://www.npr.org/2025/11/23/nx-s1-5615222/consumer-spending-is-the-u-s-economys-main-driver-heres-how-its-doing); See also U.S. Bureau of Economic Analysis data showing personal consumption expenditures at approximately 69% of GDP.
11
15 comments
THE AI LIE: Why Corporate Leaders are Gaslighting the American Worker.
powered by
skool.com/aititus-2906
Build your AI business before AGI changes everything. Real income, real freedom. I'll show you what actually works. Let's doooo this!
Suggested communities
Powered by