For fifteen years, Julian Pintat has been translating technical documents from English to German. manuals for medical equipment. filings for pharmaceuticals. Operating procedures for oil rigs. the kind of high-stakes, specialized work that calls for years of domain expertise and cannot be replicated by a general tool. His work didn’t vanish overnight when AI emerged, first with Google Translate, then DeepL, and finally with the full wave of large language models. It evolved into something more depressing: fixing AI’s errors. The translation software interpreted “scale,” which refers to mineral buildup on equipment, as both a musical scale and a weighing device on a recent oil rig manual. It frequently takes more time to correct that kind of fundamental mistake, which now accounts for about 95% of his work, than it does to translate the document from scratch. His earnings have been cut in half. He has permanently postponed getting married and starting a family. He says, “I’m the canary in the coal mine,” and it’s difficult to disagree.
In corporate America, white-collar jobs are exhibiting the pattern Pintat outlines, albeit to differing degrees. The most outspoken opponents and supporters of AI don’t usually aim for a dramatic, overnight disruption. Compared to that, it is quieter and, in some ways, more consequential. Two-fifths of global business executives stated they had already eliminated entry-level positions as a result of AI-enabled efficiencies, according to a Fast Company report published in February 2026. Press releases and layoff notices are not being used to announce the disappearing positions.
They’re just not being replenished. There are fewer interns in the class. The number of junior analysts is unchanged. The January opening for an entry-level coding position has evolved into a senior position with a job description mentioning proficiency with AI tools.
| Category | Details |
|---|---|
| The Scale of the Shift | |
| Entry-Level Jobs Reduced | Two-fifths of global business leaders have already reduced or eliminated entry-level roles due to AI-driven efficiencies, per a February 2026 Fast Company report |
| Workers at Risk | Brookings Institution research found more than 30% of all US workers could see at least 50% of their job’s tasks disrupted by generative AI — disproportionately concentrated in entry-level roles |
| Roles Most Affected | Junior lawyers, paralegals, entry-level coders, research analysts, junior underwriters, customer service reps, translators, and sales development representatives (SDRs) |
| Corporate Case Studies | |
| AIG | CEO Peter Zaffino told TIME the firm is training AI to function as a “junior underwriter” handling bulk underwriting tasks — experienced practitioners take over from there |
| A&O Shearman (Law) | Built an AI tool that scanned 20 years of license agreements to help a major US bank comply with European law; replaced what would have been “20 lawyers in a room” — halved project cost while taking on work it might have declined |
| Morgan Stanley | Built AI tools for meeting transcription, summarization, and knowledge retrieval — head of AI solutions Kaitlin Elliott acknowledged AI has automated “work typically given to junior staff” |
| Cohere’s North Tool | AI platform adopted by RBC (Canada’s largest bank) now handles 90% of general support tickets; co-founder Nick Frosst compares AI’s labor disruption to the industrial revolution |
| The Hidden Long-Term Risk | |
| The Pipeline Problem | Harvard Business Review warned in September 2025 that eliminating entry-level roles is short-sighted — these jobs are how future senior leaders develop judgment, make mistakes, and learn through doing |
| Key Question | “How will we have senior people in 5–7 years if we don’t have junior people today?” — posed by talent strategist John Vlastelica; no organization currently has a clean answer |
| AI ROI Reality Check | MIT report (August 2025) found 95% of AI pilots failing to deliver a return on investment; a Berkeley METR study of 16 experienced developers found they were 19% slower using AI coding assistants despite estimating they’d be 20% faster |
| Freelancer Impact | Translators like Julian Pintat report 95% of work now involves correcting AI errors — income halved; life plans deferred. “I’m the canary in the coal mine,” he told TIME |
The number of corporate case studies is growing. In 2025, AIG CEO Peter Zaffino told TIME that the company is training an AI system to act as a “junior underwriter,” taking care of the majority of underwriting duties so that more seasoned professionals can concentrate on the remainder that requires judgment. In order to help a major US bank comply with European regulations, a custom AI tool at London law firm A&O Shearman scanned twenty years of license agreements.
According to partner David Wakeling, this work would have required “20 lawyers in a room, maybe some paralegals.” Project expenses were halved by the tool. Morgan Stanley developed AI systems to manage knowledge retrieval, meeting transcription, and summarization throughout the company. The head of the bank’s generative AI solutions, Kaitlin Elliott, openly acknowledged that AI has automated tasks that were previously assigned to junior employees, pointing out that younger workers are now expected to make up for this by becoming fluent in AI. 90% of general support tickets are now handled by the Canadian AI company Cohere’s North platform, which RBC adopted. Although human operators are still involved, the amount of work requiring junior staff has significantly decreased.
The majority of these announcements lack an honest accounting of what is lost when the bottom of the talent pipeline is cut. Amy Edmondson and Tomas Chamorro-Premuzic argued in a September 2025 article in the Harvard Business Review that eliminating entry-level positions is truly shortsighted, regardless of any short-term efficiency gains. As a pragmatic organizational argument rather than an ethical one, though the latter does exist. Every senior employee used to be a junior employee. They learned by making mistakes, getting feedback, trying again, observing more seasoned coworkers, making embarrassing mistakes in low-stakes situations, and progressively gaining judgment that could not be instilled through a training program. The junior role must exist in order to complete that process, which takes years. You lose more than just the entry-level headcount if you skip it. The development path that generates the people you’ll require in five to ten years is hollowed out. “How will we have senior people in 5–7 years if we don’t have junior people today?” asked talent strategist John Vlastelica. As of right now, no organization has a satisfactory response.
The question of whether the efficiency gains motivating these choices are as genuine as they seem is another. 95% of AI pilots are not yielding a return on investment, according to an August 2025 MIT report. A group of seasoned developers were 19% slower when using AI coding assistants, despite their estimation that the tools made them 20% faster, according to a preliminary study from Berkeley’s METR research group. The discrepancy between perceived and actual performance is substantial, and it is probably more pronounced in companies that reduce the number of junior employees based more on theoretical AI capability than on actual output. It turns out that when you remove the human who was previously doing the work, speed is not the only thing that slows down. Institutional knowledge, context, and accuracy all slow down.
Observing this from the outside, it seems like many businesses are making a choice that won’t be fully apparent for a number of years. On a CFO’s spreadsheet, the short-term math makes sense: replace a $60,000 annual research analyst with a subscription to an AI tool. It is much more difficult to model the long-term math, which includes depleted pipelines, gaps in institutional knowledge, senior employees who are overworked, and a generation of workers who did not have the early career experiences that shaped their predecessors. That’s not to say it won’t come. It simply indicates that it will arrive following the announcement of the efficiency gains during the quarterly earnings call.
Co-founder of Cohere Nick Frosst likened the impending industrial revolution to a disruption that is ultimately handled by governments and labor unions rather than by specific businesses or employees. Regarding the scale, he is most likely correct. He omitted to mention that those who are just starting out in their careers are the ones who are most likely to be negatively impacted by such significant shifts; they are the ones who needed a place to start but had it silently taken away. Pintat described the canary as a warning system. Who is listening is the question.
