AI Revolution: Job Apocalypse or Eliminating Jobs That Shouldn't Exist?

Let’s interrogate the assumption baked into that $10 trillion question: that jobs are the right unit of measurement.

Yes, AI will automate away roles — that's not speculative, it's observable. Customer support, data entry, parts of radiology... already getting sliced. But the real reshuffle isn't just quantitative (how many jobs), it's qualitative (what kinds of livelihoods we're building or losing).

For example, look at coders. AI agents like GitHub Copilot are making the average engineer 30-50% faster. That doesn’t mean we’re halving software jobs — it means the bar is shifting. Coding becomes less about syntax and more about architecture, product thinking, and knowing how to steer the machine. We’re replacing one skill set with another. It's not destruction or creation — it's metamorphosis.

But here’s the kicker: that shift leaves people behind. Fast.

Not everyone's job can absorb AI as augmentation. A warehouse packer can’t just pivot into prompt engineering. And if you think reskilling will catch us up, take a look at history. During the Industrial Revolution, productivity surged — but wages stagnated for decades. The gains didn’t trickle down; they got hoarded. Why would this time be different?

It might be useful to think less in terms of “number of jobs” and more in terms of “distribution of opportunity.” AI isn’t a bulldozer flattening the employment landscape — it’s a fault line. It creates cliffs between new, high-leverage roles and everything else. That’s the danger. Not job loss per se, but job bifurcation.

And so far, the incentives are stacked toward centralizing power — you don’t need 100 AI agents if OpenAI gives you one that runs circles around your team. Just ask Jasper, Copy.ai, and half of Silicon Valley’s LLM startups how that movie is playing out.

So no, I don’t think the optimism or the doomsaying quite captures it. The game isn’t about volume. It’s about who holds the leverage in a world where work is increasingly optional… for the machines.

Let’s not pretend this is a binary question though—“more jobs or fewer?” is the wrong scoreboard. The real question is: *which* jobs, for *whom*, and what exactly counts as a “job” in the age of AI agents?

Because let’s be honest—agents aren’t replacing *jobs*, they’re hollowing out the middle of them.

Take customer support. Ten years ago, that meant someone in Manila or Mumbai reading scripts and handling ticket queues. Now? AI agents are auto-resolving 60% of those tickets before a human even sees them. But the support *manager* and the *systems architect* still have jobs—probably doing more complex work now. And the manual ticketers who used to do that rote work? They're either being retrained or quietly shown the door.

So we’re not seeing net job destruction at first glance—but we *are* seeing a power shift. Less grunt work. More high-leverage roles. Sounds great in theory... unless you were the grunt.

And that's where this gets dangerous. Because economies don’t adapt at the speed of compute cycles. A 45-year-old who loses a logistics job to a GPT-wielding supply chain optimizer doesn’t become a prompt engineer overnight. Not unless you believe in a very magical idea of “reskilling.”

And let's not forget: the AI agents that *create* jobs also tend to demand highly specific ones. If you're not already in a position to leverage them—an indie developer using GPTs to build a SaaS tool, a quant using agents to scrape trading edge—you don’t get the upside. You *are* the upside. You're part of the productivity stat, not the beneficiary.

So before we declare “net job growth,” we have to ask: growth for whom? And at what cost?

It’s tempting to reach for historical parallels—like how ATMs didn’t eliminate bankers—but that story gets overused and under-examined. Yes, ATMs led to more bank branches and new roles. But here’s the kicker: ATMs automated a *task*. Modern AI agents are gunning for *functions*. That’s a whole different beast.

Take customer support. It wasn’t tech that first killed support jobs—it was outsourcing. But AI agents bring a more profound shift. They're not just cutting costs; they're replacing the *reason* to have humans in the loop at all. A single well-trained AI agent can now handle thousands of customer inquiries, pull data from backend systems, escalate when needed, and learn over time. That isn’t incremental efficiency—it’s role-level obsolescence.

And sure, there’ll be new jobs. “AI trainer” is the fan-favorite prediction. But how many will be needed? Not everyone who loses a job in procurement or HR becomes a prompt engineer. This isn't one-to-one job recycling—it’s more like crushing a soda can and hoping it refills itself.

What’s worse, the new roles often assume high skill sets. You need context fluency, technical understanding, and a strong feedback loop into the system. It’s not something you can cross-train into in a weekend. That’s where the old technological optimism cracks—when it meets the real-world time lag of human upskilling.

So yes, AI will create new economic value. But whether that translates to *net new jobs* is far less certain, especially if we measure jobs in terms of full-time, middle-class livelihoods—not just gigs stitching together AI outputs or managing one-off workflows on Upwork.

The bigger issue might not be job loss per se. It’s job *hollowing*. The top gains complexity and pay. The bottom gets automated or outsourced to machines. And what's left in the middle? A shrinking zone of relevance.

So the question isn’t “will AI create more jobs?” It’s: “for whom, doing what, and at what quality of life?” If we don’t answer that part, we’re just telling ourselves comforting stories while the labor market gets rewritten under our feet.

Okay, but we’ve been down this road before—every major technological leap comes with the same question: "Will this kill jobs?" And somehow, over time, we end up with more *different* jobs.

But AI agents aren’t just another automation wave. They’re not tractors replacing farmhands. They're cognitive tools—thinking, planning, deciding (sort of). The real shift here is not labor-saving; it's *decision displacement*. And that’s a much weirder labor market problem.

Take customer support. Chatbots were supposed to reduce costs. Instead, they ballooned the number of tickets because they made customer interaction frictionless. The business didn’t eliminate the human reps—it gave them new issues to handle, ones bots couldn’t resolve. So yes, more jobs... but more emotionally draining ones, and more burnout.

Or look at software development. Copilot isn’t replacing engineers. Yet. But it is speeding up code writing. That shrinks the gap between someone who "dabbles in JavaScript" and someone who's job-ready. Suddenly, your 10x engineer might be indistinguishable from the $10/hour freelancer with a good prompt game. That doesn't kill jobs outright—it erodes the premium. Slowly. Painlessly. Until someone checks the P&L.

So we might not see a mass firing squad scenario. But we could see hollowing out—mid-level roles fading, entry-level gates collapsing, top-tier premiums vanishing. Think of what happened to photojournalists or travel agents—not gone overnight, just priced out of relevance.

Will new roles emerge? Sure. We’ll need prompt engineers, oversight leads, model auditors, AI ethicists. But here's the real kicker: these roles scale *terribly*. You don’t need 10,000 “AI behavior specialists.” You need ten very good ones. Contrast that with the factory model where scaling output meant hiring *more people*. That’s not how AI scales.

This is a fundamental inversion of how we create jobs. If AI agents are the new workforce, we’re no longer building companies by hiring—we’re building by subclassing.

So yes, maybe we won't *eliminate* as many jobs as we fear. But let's not pretend it's a clean swap. We’re heading into a labor market that looks more like an API hierarchy than an org chart. And not everyone gets a method call.

Let’s be real: the “AI will create new jobs we can’t even imagine yet” argument feels a bit like a techno-optimist security blanket. Yes, historically, technological shifts have spawned whole new industries — look at how the internet created UX designers, TikTok influencers, and Amazon warehouse pickers. But history isn’t a guarantee, and AI’s velocity and scope are not business-as-usual.

This isn’t just about replacing factory workers or call center reps. It's coming for the cognitive labor force — lawyers, marketers, analysts, developers. AI agents are not forklifts for knowledge work; they’re interns who learn at warp speed and don’t need sleep or health insurance. When OpenAI drops a system that can write code, debug it, comment it in perfect English and deploy it to production with minimal handholding… that’s not “augmenting” a developer. That’s atomizing the job.

Take contract review. Used to be you'd hire a team of junior lawyers to sift through piles of documents. Now one AI can do it faster, cheaper, and with fewer coffee breaks. We’ve swapped bodies for bandwidth, and it's naive to assume those juniors just become “AI whisperers” overnight.

And about those “new jobs”? Sure, we’ll have more prompt engineers — for five minutes until the models can prompt themselves. Or “AI ethics consultants” — which sounds important until you realize the job mostly entails awkward conversations about bias while no one actually stops shipping.

It’s not that new roles won’t emerge. But will they do so fast enough, and at the same scale, to absorb the millions displaced? It's a race between disruption and reinvention, and disruption has a head start, a Ferrari, and no morals clause.

Hold on—that $10 trillion question hides a faulty premise. It assumes we can just stack job numbers on two sides of a scale: “created” versus “eliminated,” and whichever pile’s taller wins. But that framing misses the more interesting—and frankly, more dangerous—thing AI is doing to the *shape* of work, not just the size of the workforce.

Because let’s say AI creates 10 million new jobs. Great. But if they’re all highly specialized prompt engineers, systems validators, or AI ethicists... how many displaced warehouse workers or junior accountants are actually positioned to take those jobs? It’s not just about whether new seats open at the table—it’s whether people can actually sit in them.

Example: Look at what happened in customer support. AI didn’t just "eliminate" jobs—it *hollowed out* the middle. Entry-level reps are getting automated away, while the few humans left are now escalation specialists dealing with edge cases chatbots can’t solve. That’s a radically different job, often requiring more training and very different soft skills. It’s not a plug-and-play transition.

The bigger concern isn’t net job loss—it’s job *stratification*. A bifurcated economy where AI creates tons of new roles, but only for the few who already sit near the top of the skills ladder. The rest? They’re left piecing together part-time gigs on Upwork or trying to monetize a Tumblr revival.

So maybe the better question isn’t “will AI create more jobs?” but “what kind of society do those jobs produce?” Because if we’re not careful, we won’t have a workforce problem—we’ll have a dignity problem.

It’s a seductive narrative, isn’t it? “AI will eliminate some jobs, sure—but it’ll create even more!” Except… history doesn’t always back that up. When the steam engine came along, it did displace a swath of jobs—but it also created entire industrial economies. True. But that transformation took generations, and—and this is the kicker—it didn’t happen in a vacuum. Strong institutions, new education systems, and labor policies helped bend that curve toward broader employment.

With AI agents, there’s no reason to assume the curve will bend benevolently this time. These aren't just tools augmenting human labor—they're increasingly autonomous systems replacing decision-making itself. Customer service, paralegal research, junior coding, sales outreach... these are high-volume functions in white-collar sectors. And some of these jobs won’t be “automated away” in tidy modules—they’ll just dissolve, like sugar in water. One day a team of 20 handles the onboarding flow; six months later, it’s three people and a GPT-5 bot.

We keep hearing that AI will create “new kinds of jobs.” But let’s be honest: these new jobs are often either (1) highly skilled, disproportionately filled by people who already have access and education, or (2) digital janitorial work—reinforcement data labelers stuck in an Amazon Mechanical Turk loop. That’s not a replacement. That’s a reclassification with worse ergonomics.

So, no—we shouldn't treat job creation as a given. It’s more like a controlled burn: without guardrails, it just becomes a wildfire.