'AI will take my job' — but investors are betting on office workers, not judges

Fear that "AI will take my job" is growing. Experts have long warned that highly skilled professions — doctors, lawyers, judges — face the greatest threat from artificial intelligence.

Yet the investors actually funding AI startups had little interest in building products aimed at judges or surgeons. In practice, the money flowed toward office workers, marketers and data analysts. What is technically possible and what actually attracts capital, it turns out, are two very different things.

A research team led by Dr. Fenoaltea at Italy's Enrico Fermi Research Center published those findings in PNAS Nexus, Volume 5, Issue 6, a journal of the US National Academy of Sciences.

Previous research assessed AI's threat to jobs by measuring technical feasibility — experts scored whether AI could read medical scans, analyze case law or reconcile accounts.

That approach has a critical blind spot: companies deciding whether to adopt AI do not look at technical feasibility alone.

The research team took a different approach, looking instead at where investors had actually put their money.

The team conducted a comprehensive analysis of 958 AI startups backed by Y Combinator, the world's most prestigious startup accelerator and the organization behind Airbnb, Dropbox and Twitch. Y Combinator's acceptance rate is below 1 percent — only a handful of thousands of applicants make the cut. Clearing that bar means professional investors have already vetted a venture's market potential and profitability.

The researchers used AI to map each startup's product against the occupations it targeted, producing what they called an investment-based "real threat index" for each job category.

General office workers topped the threat index. Startups are already flooding the market with products that automate repetitive tasks — document processing, database management, scheduling and phone handling.

Such work follows fixed rules and is forgiving of errors, making it an attractive market for investors. Data scientists and market research specialists also ranked among the high-risk occupations.

Judges were a different story. Even though AI can theoretically draft rulings, investors steered clear of the sector.

The reason is straightforward. If an AI system tasked with decisions that strip someone of their freedom or determine their assets goes wrong, the legal and social fallout would be severe. In the United States, controversy over algorithmic bias in criminal sentencing is already fierce. Some domains, the research suggests, are technically within AI's reach but socially off-limits.

Pediatric surgeons fell into the same category. A misjudgment by AI during an operation could cause irreversible harm, and no startup was willing to enter the market bearing that risk.

The contrast between database administrators and lawyers illustrates the gap most sharply.

Both occupations demand similar levels of information organization and logical reasoning, so their scores under a purely technical-feasibility framework are nearly identical. Yet the investment-based threat index stood at 0.8 for database administrators and just 0.05 for lawyers — the same underlying skills, but vastly different levels of AI investment aimed at replacing them.

The research team divided the occupations that investors avoided into two categories.

The first covers jobs where a single mistake carries catastrophic consequences — surgeons being the prime example. No matter how capable the technology becomes, investors are reluctant to enter fields where the margin for error is vanishingly small. The second covers jobs where automation itself triggers social controversy, with judges as the clearest case.

The gap was equally pronounced across industries.

Information and communications, finance and insurance, and professional services faced the greatest exposure. Construction, mining, and agriculture and forestry registered the lowest threat levels. AI software operates inside computers, which is why physically demanding, on-site work has remained relatively safe — until now.

However, the research team warned that combining AI software with robotic hardware could fundamentally alter that picture.

When the team isolated AI startups developing robotics technology and ran the same analysis, many on-site occupations that had appeared safe in the original study showed high threat levels.

"The combination of AI and robotics could significantly expand the scope of automation to include manufacturing and production environments," the team warned.

DOI: 10.1093/pnasnexus/pgag185

Enrico Maria Fenoaltea, Dario Mazzilli, Aurelio Patelli, Angelica Sbardella, Andrea Tacchella, Andrea Zaccaria, Marco Trombetti, Luciano Pietronero, "Follow the money: A startup-based measure of AI exposure across occupations, industries, and regions," PNAS Nexus, Volume 5, Issue 6, June 2026, pgag185.