Companies Are Firing the Experts Their AI Needs to Learn From

Oracle employee Jill spent months training an AI system on her specialized workflow. Then the company fired her.

She wasn't alone. According to Time's investigation, Oracle systematically had workers document their processes for AI systems before eliminating their positions. The pattern reveals a contradiction at the heart of corporate AI strategy: companies are removing the very experts their AI systems need to keep improving.

"For AI systems to keep improving in knowledge work, they need either a reliable mechanism for autonomous self-improvement or human evaluators capable of catching errors and generating high-quality feedback," VentureBeat's analysis notes. "The industry has invested enormously in the first. It's giving almost no thought to what's happening to the second."

The Numbers Tell a Different Story

Gartner surveyed 350 global businesses with revenues above $1 billion. All were piloting or deploying intelligent automation. The findings contradict the boardroom narrative about AI efficiency.

The research shows AI-driven layoffs rarely deliver promised returns. Companies discover too late that their AI systems can't actually perform the nuanced tasks their human experts handled. The Register reports these firms are finding "the strategy isn't working."

The problem runs deeper than implementation hiccups. When companies fire their domain experts, they lose the people who catch AI errors, provide training data, and handle edge cases the models can't process.

The Knowledge Death Spiral

Shomron Jacob, head of Applied Machine Learning at iterate.ai, warns employers are "placing their businesses at strategic risk by cutting headcount before being sure AI can perform the tasks they need it to."

The risk compounds over time. Today's AI models train on data created by human experts. When those experts disappear, who generates tomorrow's training data? Who validates the AI's outputs in specialized domains?

VentureBeat calls this the "expert replacement paradox." Companies assume AI performance will keep improving after they fire the humans who made that improvement possible. They're betting on autonomous self-improvement that doesn't yet exist.

Oracle's Playbook Spreads

The Oracle pattern has spread across tech. Employees spend months documenting their workflows, training AI systems, creating detailed process maps. Then they're let go.

"Everyone's a line on a spreadsheet," one Oracle worker told Time. The human cost is obvious. The business cost takes longer to surface.

Gartner's study found companies rushing to cut costs through AI automation often face unexpected expenses later. They hire consultants to fix broken processes. They bring back contractors at premium rates. They lose institutional knowledge that no model can replicate.

The Feedback Loop Problem

AI systems need continuous feedback to improve. In customer service, that means experts who know when the chatbot gave bad advice. In legal work, it means lawyers who catch when the AI misinterprets precedent. In engineering, it means developers who spot flawed code.

Remove those experts, and AI performance degrades. Errors compound. Edge cases multiply. The system that looked impressive in demos fails in production.

Diginomica warns that companies cutting staff before confirming AI capabilities face strategic risk. The eighteen-month corporate obsession with AI efficiency has ignored a basic requirement: someone needs to keep teaching the teacher.

The companies firing their experts today are betting those same experts won't be needed tomorrow. Gartner's research suggests they're wrong.