24/05/2026
The AI Job Apocalypse Is the Wrong Conversation. Work Itself Is Evolving.
The headline everyone is debating
On 22 May 2026, Goldman Sachs CEO David Solomon published a guest essay in The New York Times — picked up the same day by Forbes — arguing that fears of AI-driven mass unemployment are “overblown”. He pointed to a century of U.S. labour absorbing electrification, computing, and the internet, and noted that data-centre construction alone has added more than 200,000 jobs since 2022.
He is not wrong. But he is answering the wrong question.
The debate is stuck on a binary: will AI destroy jobs, or will it create them? That framing is comfortable for CEOs, terrifying for employees, and useless for leaders who actually have to make decisions in the next 24 months.
The real question is sharper: What does a “job” even mean inside your organisation a year from now?
What Solomon’s own example actually reveals
Read his essay closely. Solomon says Goldman will need fewer people for regulatory reporting, client onboarding, and document review — but more client-facing bankers, traders, and asset managers. Elsewhere, Goldman’s COO has openly described deploying AI “robots” on the firm’s internal “human assembly line” while pledging no mass layoffs.
That is not job preservation. That is work recomposition. The role title on the business card may stay the same. The tasks inside it will not.
Goldman’s own economists make the friction visible: they estimate AI could displace 1–4 million U.S. jobs annually, and warned earlier this year that AI-related layoffs could push U.S. unemployment toward 4.5% by end-2026. Both things are simultaneously true: net job creation continues, and individual careers are being rewired underneath the headline numbers. Stanford research cited in Solomon’s own essay shows entry-level roles in highly automatable fields have already fallen 16% versus less-exposed roles.
The apocalypse isn’t coming. The recomposition already arrived.
Why “evolution of work” is the right frame
At TCPS, we have spent 13 years and 70+ enterprise engagements studying what actually happens inside organisations when the ground shifts. Three patterns are now unmistakable:
• The unit of work is shrinking from “role” to “task.” AI doesn’t replace accountants; it replaces reconciliation. It doesn’t replace researchers; it replaces literature scans. Leaders who plan headcount instead of task portfolios are solving last decade’s problem.
• Human intelligence is being repriced, not removed. Judgment, sense-making, relationship trust, cultural fluency, and ethical reasoning are now the scarce inputs. Goldman keeps client-facing bankers because clients still buy trust, not throughput.
• Culture is the bottleneck, not technology. AI adoption stalls not because models are weak but because organisations cannot redesign workflows, incentives, and identity fast enough to absorb them. This is a culture problem dressed up as a tech problem.
What this means for Singapore leaders specifically?
Singapore’s workforce is highly white-collar, highly exposed, and highly trainable — exactly the profile Solomon flags as facing the heaviest task automation in accounting, banking, and law. The national instinct will be to retrain into new roles. That instinct is half-right.
The other half: organisations must redesign the role itself before sending anyone to be retrained. Otherwise we will produce highly upskilled workers stepping back into job descriptions written for 2019.
Our position
We believe the next 24 months will separate two kinds of organisations:
1. Those that treat AI as a headcount question — and end up with demoralised teams, hollowed-out career ladders, and a culture that quietly stops trying.
2. Those that treat AI as a human intelligence question — and rebuild work around what humans uniquely contribute when machines handle the rest.
This is the category we have been building toward since 2013. Our HEARyou qualitative research suite was designed to surface what employees actually think, feel, and fear — the signals that dashboards miss. Our CULT framework was built to translate those signals into culture-based redesign of how work gets done. Together, they answer the question Solomon’s essay sidesteps: how do you re-author work itself, not just retrain the worker?
The one sentence to take away
The AI apocalypse is overblown. The AI recomposition is underway. Leaders who confuse the two will spend 2026 defending headcount; leaders who understand the difference will spend 2026 redesigning what their people are actually for.
---
Curious to see what recomposition looks like inside your organisation?
TCPS runs structured HEARyou diagnostics and CULT redesign engagements with enterprise and government clients across Singapore. Contact us to start a conversation.
David Solomon acknowledged advancements in artificial intelligence have eliminated jobs in some industries but “may lead to job growth in others.”