The Four Phases of a Potential AI Jobs Apocalypse

The Right Question about AI and Jobs

Every day, another frustrated job seeker, computer science degree in hand, sends resumes into an abyss. Is this just a soft patch in the economy-- a time for companies to wait out some of the current uncertainty? Or a structural, AI-driven shift? A few months ago, Anthropic founder Dario Amodei suggested the latter. He said AI would take out half of entry-level white collar jobs in the next three to five years.

All the AI luminaries weighed in on his prediction. They tried to offer nuance, but the world loves a binary question.

Will this technology destroy white collar jobs?

People with deep expertise offered diametrically opposed answers. We form our camps, and move on. Or don't. This same discussion comes up again and again. Nothing really happens. Nothing but the inexorable march toward better, faster, stronger, more capable AI.

Maybe we should reframe the question. Maybe that would help us have a more informed discussion about what we really want. Maybe that would lead to some action.

Here's an attempt:

If AI advances don't hit some ceiling (a big and important IF), how might AI (and eventually Super Intelligence) re-shape the labor market if we fail to intervene?

The Folly of Prediction

I've spent a lifetime in the jobs business. I've seen bubbles come and go (telecom expansion, Y2K, dotcom, mortgage) and I've seen how automation and offshoring have re-shaped industries.

I remember Big 3 Automakers telling analysts about all the layoffs they were executing the same day they were calling us for software engineers, supply chain talent, and other key skills. They pumped up their cost-cutting while downplaying the hiring. They were talking to Wall Street, not Main Street.

In 1985, the US had 485,000 bank tellers. Twenty-two years later, with ATMs in every supermarket, we had 600,000 bank tellers. Technology had lowered costs and expanded reach. But that's not always the case. Town criers and buggy whip manufacturers didn't fare so well.

All to say, the prediction business is tough. Complex systems don't lend themselves to accurate forecasts. Too many variables. Too many players.

Still, I've developed a working hypothesis based on what I know about past technological disruption, the state of the labor market, our capitalist operating system, and the potential broad-based power of AI.

It will be wrong.

It's hard to fathom what true AGI or Super Intelligence would mean. But that doesn't mean we shouldn't try.

🌪 Phase One: The Hollowing (WE ARE HERE)

Winners: Leaders of AI Experiments

🔻 Endangered Species: Entry-Level Office Workers

📉 What Transpires: Companies allocate investment to labor-replacing AI. They slow or freeze early-career hiring, and promise the market they're on the leading edge of this revolution. Inside, though, they worry they're not keeping pace and operate from a position of fear. The effect is a hollowing out of their traditional hierarchy.

Broadly speaking, any leverage labor had during the Great Resignation has been wiped out in the white collar arena. While recent college grads feel the most immediate impact, mid-career professionals start to question their future. Their employers don't have great answers for them because they don't have great answers. They understand their legacy business model could be under attack, but they're not confident how they need to transform. Many focus time and attention on automating their existing process, which will ultimately prove a fool's errand and race to the bottom.

_________________

Key Question: Will Agentic AI break through and prove to be a General Purpose Technology (moving from Answers to Action) and will employers effectively deploy the technology as more than point solutions?

If so, brace for The Culling…

_________________

⚙️ Phase Two: The Culling

Winners: Domain Experts Who Leverage AI to Amplify Their Impact

🔻 Endangered Species: Workers Resistant to Change

📉 What's Happening: Companies have figured out new, agentic models. They also recognize the need for constant experimentation. They have an unquenchable thirst for workers who have a fluency with the tools (both people who can manage agents and people who can test new models). Those who do not demonstrate such fluency will be exited quickly. Said another way, "ten years experience" gets displaced by a twenty-something ready to manage 100 agents.

Upstart competitors challenge incumbents. Legacy companies feel emboldened, but still largely operate out of fear. Big layoff announcements happen on a daily basis, but behind the scenes, there's strong demand for talent who can leverage the agentic workforce.

Markets look more and more like winner-take-all competitions as new platforms emerge. Many companies compete to orchestrate those platforms; others run the wrong race, accelerating to the bottom, spending on automation with ROI calculations that never prove out.

_________________

Key Question: Will AI adoption break from past technically-charged revolutions with intelligent machines taking on higher-level jobs (as opposed to freeing up humans to do higher level jobs)?

If so, ready yourself for The Shakeout…

_________________

🔀 Phase Three: The Shakeout

Winners: AI-Native Organizations and Platform Orchestrators

🔻 Endangered Species: Legacy Business Models

📉 What's Happening: Pre-AI companies struggle to reconfigure themselves quickly enough to compete. AI-natives emerge, leveraging a largely agentic workforce and network effects to become the dominant platforms across most industries.

Younger workers will see little advantage to joining large enterprises that do not control the platforms. Industries reform as decentralized, platform-driven ecosystems; rigid organizational hierarchies and defined job titles collapse. The Capital vs. Labor tension shifts again, but not to traditional labor. Entrepreneurs will see a Cambrian Explosion of opportunity; capital will flow to them. A Pareto distribution of winners will emerge.

Many workers will be left behind. They will look for employment in non-office environments (health care, service workers, construction, trades).

_________________

Key Question: Will we achieve Super Intelligence (or AGI or whatever the in vogue term is) leading to a world of abundance with near-zero-cost access to goods and services and a largely machine-run economy? 

If so, time to confront Existential Drift…

_________________

🧭 Phase Four: Existential Drift

Winners: Tech Oligarchs

🔻 Endangered Species: Legacy Institutions, Work-Related Purpose and Dignity

📉 What's Happening: AI abundance makes goods and services cheaper and more accessible. As margins shrink to almost nothing, many historic private domains become utilities. Machines take on the lion's share of work. Labor loses all leverage, and we enter an era where we need to re-imagine identity, meaning, and the social fabric once provided by jobs and workplaces.

The prior phases will feel like pebbles in the ocean compared to the tsunami of change that happens during this phase. Government and social institutions prove inadequate for the moment. The concept of Nation State (a recent construct) comes under fire. Global companies can cherry-pick the most favorable regulatory environment. The Tech Oligarchs seize even more power as they deploy self-improving AI to dominate industries and people.

Considerations

While I would submit that the future's cloudy (and gets cloudier as we move toward Phases Three and Four), this framework invites a more nuanced discussion. First, we need to ask what agency we have or want. The current sentiment seems to be: we're on an inexorable march toward some future that will be defined by technical advances and fueled by geopolitical competition.

Are we okay with outsourcing our future to a handful of tech billionaires and the politicians they sponsor?

Beyond that, we need to start a grown-up conversation about how we prepare for a radically different future. While I have a low confidence interval that we'll get to Phase Four and I haven't attached a timeframe to the phases, I wouldn't advocate an ostrich strategy. We should be coalescing around a set of assumptions and then constantly updating those assumptions as more information comes available.

I won't endeavor to answer the myriad of questions we'll need to consider, but I'll at least raise a few that give me some pause.

Consumption-Based Economy: taking cost out of goods and services makes those goods and services accessible to more consumers. But what happens if those potential consumers are, at the same time, seeing their income shrink? Henry Ford made Model-T's available while he raised factory worker's wages, helping to build a middle class of consumers. We've seen the flip-side. Automation, offshoring, and the growth of Wal-Mart hollowed out the middle class in what's now called the rust belt. Those communities did not thrive even with the access to cheaper goods and services. What makes us think this will play out differently?

Inequality and the wealth redistribution challenge: To hear the accelerationists talk, we won't have to worry about Universal Basic Income; we'll have Universal High Income. We're moving to a world of such abundance that the rising tide will benefit everyone. Color me skeptical. Do we have an example where the winners in a winner-take-all economy have willingly redistributed wealth? And do we really believe that the tax-evading libertarians in Silicon Valley will have some epiphany that causes them to advocate for progressive redistribution policies? Still waiting for evidence that our current and future oligarchs understand the concept of "enough."

Upskilling/ Reskilling: The easiest answer is to make sure those affected have ample opportunity to reskill for more in-demand opportunities. Our recent history suggests that it’s the exception, not the rule, that the laid off production worker in Flint will take a coding class and find employment. The reality: our economy does a remarkably good job of allocating workers to in-demand areas, but this happens over decades, not years or months. We're going to have to accelerate these efforts, radically change higher education and/ or find other means of dealing with affected workers. I suspect most efforts will prove inadequate, but I worry we're not focused enough in this area.

The Relationship between Purpose and Work: Most of the air time in these discussion seems to go to engineers and economists. Politicians like to weigh in as well. We don't hear enough from philosophers and psychologists. The real problems, I suspect, will be related to dignity and purpose as much as income redistribution. We would all be wise to re-read Victor Frankl's Man's Search for Meaning. Or maybe ask Chat GPT about it.

Government Fighting the Last War (or the Wrong War): The tech companies are in full regulatory capture mode. They have enough money and our political system has been bent enough to embrace their "donations" that it's hard to see how our elected officials will play an active role in shaping our future. They haven't effectively dealt with decade-old problems related to social media's harmful effects. Do we think they'll be fighting the right war in this respect? And will our polarized, sclerotic system be nimble enough even if its key players weren't in bed with the tech industry?

Call to Action

AI is already reshaping our world. If scaling laws hold, we ain't seen nothin' yet. Each of us has agency and opportunity when it comes to deciding how we want to leverage AI to shape our future.

We should demand transparency from our leaders. We need to push for sensible regulation.

We should invite non-tech voices into the AI debate.

And we should recognize that framing AI as a win-lose race for world domination brings undue risk. We need sensible governance and shared goals that ensure humans flourish.

It's time we exercise our voice and our agency. Our children are counting on us.