Inevitability
My first reaction to a lot of AI predictions about the workforce is pretty similar to my reaction to most technology timeline claims: skepticism. Self-driving cars were a couple of years out for about a decade. Quantum computing has been five years away since roughly the 1980s. At some point the pattern becomes its own data point, and I’ve learned to pay as much attention to who’s making the prediction as to the prediction itself.
The specific claim that keeps coming up about AI today is ‘inevitability’. AI is inevitable. Job displacement is inevitable. Entire industries transforming beyond recognition are inevitable. I’ve sat across from executives, consultants, and product leaders who say it with complete confidence. The way it gets framed, the decision feels like it was already made somewhere upstream, and the rest of us are just figuring out where we stand.
I use AI every day. The productivity gains are real when you apply these tools to the right problems. What I keep wondering, though, is which version of “inevitable” we’re actually talking about, and who decided.
Some things about AI do feel genuinely inevitable to me. Adoption is going up and will keep going up and the underlying capabilities are improving. Businesses that figure out how to use these tools well will have real advantages over businesses that don’t.
What I don’t see as inevitable is the specific version of the story that dominated the last few years: AI is going to replace your workforce, AI will make most knowledge work obsolete. There has been lots of ‘get ahead of the wave or get flattened by it’ conversations. These versions of the story came from somewhere…but where?
Who’s been talking
If we follow those stories back in time and start looking at who was doing most of the talking, we can see the ‘inevitable’ story for what it’s worth.
The loudest voices on AI job displacement have been the people selling AI tools and services. The incentives here aren’t complicated: if you’re running an AI company or selling AI SaaS, you benefit when people believe the technology is both inevitable and existentially threatening to existing workers. That belief drives urgency and urgency drives purchasing decisions.
Reuters Institute analyzed UK media coverage of AI and found that one third of articles were based on industry sources (mostly CEOs and senior executives), six times as many as those from government. The people selling the technology are explaining what it will do to you.
Daron Acemoglu, the MIT economist who won the 2024 Nobel in economics, told Fortune in June 2026 that “about 20% of AI discourse is intellectually serious.” He’s spent years modeling actual labor market impacts. His research suggests roughly 5% of tasks can be profitably automated near-term at current costs and capabilities. McKinsey projects $4.4 trillion in annual productivity potential . Acemoglu’s modeling estimates 0.53-0.66% total factor productivity gains over ten years. The gap between those numbers is significant, and almost nobody writing about AI strategy has spent much time on the question of which one is right.
A 2025 academic paper, “Making AI Inevitable,” examines how determinism gets embedded in AI coverage. The argument is that every prior major technology transition got shaped by institutional and political choices, not fate. The “inevitable” framing strips those choices out of the story.
That’s the framing problem but this framing keeps running into reality. The evidence is starting to pile up the other way and it doesn’t look much like the vendor projections.
Companies that publicly announced they were replacing workers with AI have been hiring again. The real-world economics turned out to be different from the projections. Running AI to do the work a person used to do often costs more than the person. The models need oversight, they make mistakes that need correction and workflows that looked clean in a demo get complicated when they touch real systems.
Harvard Business Review published a study in January 2026 of over a thousand executives and found that a significant number of companies were making workforce reduction decisions based on expected AI capabilities, not demonstrated ones. They were betting on a projection, not a result.
MIT’s NANDA initiative, reported in Fortune in August 2025 , found that 95% of generative AI pilots failed to deliver measurable ROI. Gartner predicted at least 30% of generative AI pilots would be abandoned after proof of concept by end of 2025. Those numbers don’t get press releases.
There’s also a quieter mechanism at play here. Many companies aren’t actively replacing workers with AI. They’re just not backfilling when people leave. Entry-level knowledge work is disappearing without visible mass layoff events. Why hire a new analyst when AI might handle that in eighteen months? The decision gets made based on a capability that hasn’t arrived yet. That’s a different kind of harm than “the robot took my job,” and it’s almost entirely absent from the coverage.
The other direction
The tools do work in specific, well-defined contexts where the economics make sense. The sweeping job replacement timeline has been wrong so far. And the loudest version of it came from people who needed you to believe it.
So far I’ve been pointing the skepticism at the vendors. Fair is fair, the same question runs the other direction.
Workers whose jobs feel threatened have legitimate fears and every reason to amplify the risks. Unions have political incentives to resist the technology regardless of the specific evidence in specific contexts. Some of the most alarmist AI coverage comes from people whose livelihoods are genuinely at stake. That concern is well-founded but it’s also colored by self-interest in ways that make it hard to separate the genuine fear from the strategic amplification.
Acemoglu’s comment about 20% of AI discourse being intellectually serious applies in both directions. The people telling you AI will replace everyone are not being rigorous and the people telling you AI is a fraud that will never change anything are not being rigorous either.
What rigorous looks like is asking: which specific tasks, at what cost, on what timeline, and what does the actual evidence show?
Which brings me back to that word, “Inevitable.”
When someone says job displacement is inevitable, they’re making a prediction about a specific mechanism on a specific timeline at a specific cost structure. Each of those claims is empirically testable. Each of them is currently contested by serious researchers.
Asking who benefits from your belief isn’t cynicism. Any argument made by someone with a financial stake in the conclusion deserves that question. That applies to AI vendors and it applies to people whose jobs are threatened. It also applies to the consultants who get paid to advise on AI strategy, myself included.
Every major technology transition in history has been shaped by choices: institutional decisions, regulatory pressures, and economic interests. Electricity, the internet, mobile computing: none of them arrived pre-determined. The people who navigated them well stayed curious about what was actually happening rather than taking cues from whoever stood to profit.