Two major trends are colliding in the AI economy: some jobs face imminent automation risk, while others are seeing compensation explode. Nate Jones and Michael Cricggsman explore which roles are most vulnerable — and why the least likely to be replaced might surprise you.
Jobs Most (and Least) Likely to Be Disrupted
A recent Microsoft study analyzed over 200,000 Copilot chats to determine which occupations face the greatest AI disruption. The findings reveal a striking pattern: historians ranked second among jobs most likely to be impacted by AI, with interpreters and passenger attendants also appearing in the top tier.
The reason? AI makes reading primary sources dramatically easier — transforming how historians work rather than eliminating their work entirely. For passenger attendants, automation is already visible in parking garages that now operate without human staff. Flight attendants face a different reality: their role involves far more than ticket collection, requiring judgment in constrained environments at altitude.
Yet the least likely to be replaced include dredge operators, bridge and lock tenders, water treatment plant operators, pile driver operators, and certain medical technicians. These roles demand operating environments that are tightly controlled and physically challenging — conditions where robotics remains limited.
The AI Compensation Divide
The job market reveals something striking: there aren't enough AI researchers in the world, and companies are paying premium prices to secure them. Zuckerberg's offers represent just the visible tip of an iceberg hiding vast demand for specialized AI talent.
Jones distinguishes between three distinct labor markets:
First, AI researchers command roughly ten times the compensation of engineers. Some received billion-dollar liquid packages that no one accepted — a remarkable signal of how critical these roles are to building next-generation models.
Second, AI engineers see significant bumps over traditional software compensation but not the extreme premiums.
Third, skilled general engineers remain in demand because AI has made software development cheaper while raising standards for polished products. The best developers deliver 10x or 20x the efficiency of average performers.
The critical question: how do organizations measure variance within job roles and encourage extraordinary performance?
Trust Barriers in Medical AI
One of the most fascinating developments involves trust — specifically, patient trust in machine diagnosis. Radiologists have adopted AI tools that can identify pathology patterns faster than doctors with less information. Yet radiologist demand hasn't decreased; it's actually expected to rise.
The liability question remains unresolved: physicians cannot delegate legal accountability to machines. When something goes wrong in court, the doctor bears responsibility — not the software.
More compelling is the trust issue itself. In social media videos, Jones proposed that patients prefer human diagnosis even when AI tools show promising results. The response was unexpected: many viewers argued they'd prefer machine diagnosis precisely because their doctor never listens to them.
This dynamic suggests AI diagnostic tools will increasingly supplement rather than replace physician judgment — and that physicians who build trust with patients may survive automation better than those relying solely on technical accuracy.
Product Management's Technical Turn
A counter-argument emerging in tech circles: software product development roles face disruption through architect and DevOps downsizing, while business and product roles remain safe. Jones pushes back against this conventional wisdom.
The actual trend shows product management becoming more technically demanding. One observable shift involves startups hiring two PMs per engineer — a radical span reduction that allows PMs to validate ideas and manage bottlenecks rather than direct development work.
This raises an uncomfortable question: are engineers better prepared for product roles than traditional PMs? The answer remains unclear, but the boundary between technical and business functions is blurring faster than most professionals realize.