- Published 18 March 2026
THE AI DISPLACEMENT PARADOX
Two posts have been trending since yesterday and both been extremely enlightening.
Matt Shumer’s “Something Big Is Happening”, now past 70 million views on Twitter, where the HyperWrite CEO says AI replaced him at his own job and the rest of us are next.
And Connor Boyak’s Bastiat-inspired counter: we’re only seeing the jobs being destroyed while ignoring the new industries that will emerge, as they always have.
At this moment, I think Shumer is largely right about what’s happening. I also think Boyack’s framework is historically valid. But neither is asking the question that actually keeps me up at night. And that was the reason I thought of joining the discourse.
The question isn’t whether AI will displace work. It will.
The question isn’t whether new industries will emerge.
They will. The question is: what happens in the gap between the two?
The Scale Problem
Dario Amodei said at Davos 2026 that AI could eliminate 50% of entry-level white-collar jobs within 1 to 5 years. The IMF says 40% of global jobs are exposed. The WEF says 41% of employers plan workforce reductions by 2030. The IDB flags 980 million jobs at high disruption risk in the next year alone.
I run a 200 people technology sevices company. I live in this space. Shumer isn’t exaggerating. METR’s data shows AI completing tasks that take a human expert five hours, with that capability doubling every four to seven months.
But here’s what the Silicon Valley conversation misses. They talk about this as a tech-sector problem. It’s not.
India’s IT services sector employs 5.4 million people. Add BPO/KPO and you’re at millions more, precisely the cognitive, screen-based jobs that are most immediately exposed. A 30-40% disruption in India’s services economy affects more people than the entire workforce of many European countries. Multiply across the Philippines, Poland, South Africa, Brazil.
If we believe the IDB data, We’re talking about a billion people getting impacted simultaneously in 3 to 5 years. That has never happened with any innovation in human history.
And here’s what makes it structurally different from every prior revolution: when factories automated, workers moved to offices. When the internet disrupted retail, workers moved to logistics and digital services. Each time, there was an adjacent safe zone. AI doesn’t leave that gap. It’s a general-purpose substitute for cognitive work, improving across all domains simultaneously. Whatever you retrain for, it’s improving at that too. The Cycle Nobody Is Talking About
Everyone is debating jobs. I’m worried about consumption.
What happens when 50-60% of cognitive work gets displaced across multiple sectors at the same time?
Workers lose income across legal, finance, tech, healthcare, media and this is happening simultaneously. These workers are also consumers, Consumer spending is 60-70% of GDP. Demand contracts. Businesses cut costs with more AI automation. More displacement, less spending, more automation. A self-reinforcing loop. Mere feet of getting displaced contracts the demand, as Shumer was suggesting that we all should start saving.
Then comes the secondary wave. Falling revenues mean companies pull back on R&D and innovation spend. The entire ancillary services ecosystem, consulting firms, implementation partners, training providers, starts shrinking. Their employees spend less and Demand falls further.
This is a deflationary spiral. Irving Fisher described the mechanics in the 1930s. But that version was triggered by a financial shock. This one would be triggered by a labour shock, the systematic elimination of the purchasing power that keeps the economic engine running.
Some analysts call it “Japanization without stagnation”; GDP and corporate profits look healthy thanks to AI productivity, but aggregate wage growth flatlines and employment contracts. The appearance of prosperity masking a structural hollowing out.
Where the “Unseen Logic” Breaks Down
Boyack’s framework draws from Bastiat you can see jobs being destroyed, but you can’t yet see the new industries that will emerge. Trust human ingenuity, Trust markets and look at the unseen. I have been trying really hard for sometime to visualise that, have been consulting experts around the creation of new sectors but have always come up short.
I agree, historically, this has been right. The PC killed typing pools and created the digital economy. The internet created occupations nobody imagined in 1995. McKinsey estimates the PC alone created 15.8 million net new US jobs since 1980.
I respect this argument. But I see four structural problems applying it to AI:
No safe adjacent sector: Every prior revolution’s new industries required human capabilities machines couldn’t perform. But AI is capable in precisely the cognitive-creative domains where new jobs would emerge. It enters new fields simultaneously with humans, potentially faster.
The speed mismatch: New industries historically took 20 to 40 years to scale into mass employers. AI displacement is happening in 2 to 5. Human systems – education, institutions, regulation – operate on decadal timescales. What if we are not able to retrain a billion people in 18 months and find new avenues.
The compensation gap: New AI-era jobs pay well (56% wage premium per IMF) but aren’t created in sufficient numbers. The middle-class, middle-skill jobs – the backbone of consumption economies – are exactly what’s being hollowed out.
The distribution problem: New industries won’t emerge where displaced workers live. A BPO worker in Hyderabad cannot become an AI safety researcher in San Francisco. The IMF confirms: advanced economies capture disproportionate gains while emerging markets face GDP declines.
That said, at many places Boyack is right, I’m not dismissing the optimists entirely. Human creativity is unbounded. Radically lower costs will create demand we can’t yet imagine. If AI makes software development 90% cheaper, millions of previously unviable projects become possible. That’s real. Jensen Huang of Nvidia makes the point that greater productivity has historically led to more hiring, not less. And at macroeconomic timescales, this has been true.
The question – my question – is whether the transition period is survivable. Especially for countries like ours (India) that have built entire economies on the work which are the first ones caught in crosshairs. Unfortunately, at this time there is no clarity. The Three Phases Ahead
Having thought deeply about this, here’s the framework I’m working with:
Phase 1 – The Displacement Wave : This is the period where Shumer is most right. AI capabilities advance faster than organizations, education systems, or policy can adapt. 50% or more of entry-level cognitive work transforms or disappears. Corporate headcount drops even as technology investment rises. The services industry faces pressure as enterprises demand more value from fewer partners.
Phase 2 – The Trough: This is the period we all must worry about. The cumulative effect of displacement begins to overwhelm job creation from still-nascent new industries. Consumer spending contracts. Companies that over-invested in AI without clear ROI pull back. A potential deflationary cycle emerges. Governments respond with UBI experiments, retraining programs, possibly AI taxes, but these might take years to scale. This is the gap. The danger zone. This is what we all have to fight against and ensure that the next phase start earlier than expected.
Phase 3 -The New Equilibrium: This is where Boyack’s “unseen logic” starts to manifest. Radically reduced costs of cognitive work unlock new industries and forms of value genuinely unimaginable today. Human work reorganizes around deep relationships, physical-world presence, creative direction, moral judgment, care, and leadership. New compensation models emerge. The economy stabilizes – but the composition of work looks nothing like today.
Again, the critical variable is the depth and duration of Phase 2. That’s what determines whether this is a painful adjustment or a genuine crisis. That’s what should be the focus on.
Conclusion
What This Means for Those of Us Building Companies
I started this piece because these questions were troubling me. Let me end with what I’ve understood, at least for now. (It is changing on weekly basis)
If you run a technology services company, and I do, the traditional model of billable hours, staff augmentation, and headcount-based pricing is in the direct path of disruption. That’s not a 2030 problem, That’s a 2025 problem.
Enterprises are going through the most complex transformation in business history. Many war fronts are open out there and they don’t just need AI tools, they need partners who can help them redesign their operating models, rethink workforce strategies, manage the human side of the transition, and create entirely new forms of value. That work requires trust, deep relationships, organizational understanding, and judgment. These are precisely the things AI cannot replace, because they depend on human intimacy with the customer’s reality.
Both Shumer and Boyack made me think again. The displacement is real. The emergence of new possibilities is also real. But the transition between them isn’t automatic, it isn’t painless, and it isn’t going to distribute itself equitably without deliberate action.
The leaders who see both the threat and the opportunity clearly, without surrendering to either panic or comfortable historical analogies, will be the ones who shape what comes next.
