White-Collar

White-collar is a term whose name literally refers to a “white collar.” It denotes occupations in which people work at a desk: accounting, administration, law, marketing, consulting, programming. The name comes from the era when office employees wore white shirts and factory workers wore blue work clothing. This gave rise to the counterpart blue-collar for physical work in workshops, factories, or on construction sites. Those who work white-collar jobs mainly process information: reading, calculating, writing, deciding, discussing. The tool is usually a computer, and the result is a document, a figure, or a recommendation.

Why office work of all things is coming under pressure

For a long time it was assumed: machines replace muscle power, not mental work. Robots weld car doors, but they don’t write contracts. With the new language models, this expectation has flipped. Language models are programs that can read texts and generate texts themselves. That is precisely the raw material of office work.

That’s why economic news today features estimates of how many office tasks can be partially automated. Studies by major banks and research institutes often cite shares ranging from a quarter to half of all work steps in such occupations. These figures are projections, not measurements. What matters is the distinction between task and job: a job consists of many tasks, and if three of them disappear, the whole job doesn’t vanish.

For investors, though, the point is concrete. Personnel costs are the largest line item for consultancies, law firms, and insurers. If software takes over part of that, it changes the profit margins of entire industries. Conversely, it’s notable that skilled trades are hardly affected by this wave, because an AI can’t seal a pipe.

Which desk tasks software is taking over

Affected above all are tasks that are repetitive and whose output is text or numbers. An AI can summarize a draft contract, compose a form letter, analyze a spreadsheet, or write a first draft of code. These are typical entry-level tasks that newcomers used to learn their trade with.

It gets harder with anything that requires responsibility. A model can prepare an annual financial statement, but it isn’t liable for it. It can gather arguments, but it doesn’t sit in the courtroom. Tasks involving many stakeholders also remain human: negotiating, resolving conflicts, delivering bad news to a client.

In practice, this usually doesn’t result in replacement but in a shift. A team no longer writes the draft itself but reviews and corrects it. That sounds harmless, but it changes how many people are needed for the same amount of work. A common mistake here is attributing every announced job cut to AI. Often what’s behind it is simply a weak economy, and AI is the more convenient explanation.

The term in quarterly reports and job listings

In the news, you’ll usually encounter white-collar in two contexts. First, in corporate announcements: companies announce cutting administrative positions while simultaneously investing in AI. Second, in labor market reports examining whether entry-level workers in office occupations find it harder to get a job.

Product advertising also uses the term. Providers talk about “white-collar automation” or AI agents, meaning programs that carry out multiple work steps independently in sequence. These refer to tools for accounting, customer support, or research.

For you personally, the practical question is not whether an occupation will disappear. It’s which part of an occupation is routine and which part requires judgment. Those who master the latter work with the tool rather than against it. That is exactly what many of the education and training debates of recent years have been aiming at.