For about a century, two colors have done a lot of work in how we talk about jobs. Blue collar meant hands-on work. White collar meant office work. The labels were never precise, but they stuck because they gave a quick picture of what a job involved. This page traces where they came from, how the boundary between them kept moving, and why AI agents now call for a third label.
The short answer: the collar terms began as descriptions of clothing, white collar first recorded in the early 1920s and blue collar in the late 1940s, and became shorthand for two kinds of work. Each new wave of machines shifted tasks across that line. Generative AI and AI agents shift it again, this time inside office work, which is why we use the term digital collar for work that consists mainly of directing and checking automated systems.
Blue collar: the clothes of manual work
The term blue collar comes from the sturdy blue denim or chambray shirts that manual workers wore, colors that hid dirt and grease. It became the everyday name for trades, factory and other hands-on work in the United States during the twentieth century. Dictionary.com dates the first recorded use of blue-collar to 1945 to 1950, and describes it as referring to wage earners who wear work clothes or specialized clothing on the job, such as mechanics, longshoremen and miners (Dictionary.com, blue-collar).
The label came to cover factory, construction, maintenance, transport and other physical work. It also picked up assumptions about pay and education that were often wrong: many skilled trades have long paid more than routine office jobs.
White collar: the clean shirt of the office
White collar refers to the white dress shirts and detachable collars worn by office workers through much of the nineteenth and twentieth centuries. The label came to stand for clerical, administrative and management work. It is the older of the two terms: Dictionary.com dates its first recorded use to 1920 to 1925 (Dictionary.com, white-collar), a generation before blue collar.
White collar workers were a minority in farming and early industrial societies. In industrialized countries they became the majority, as economies modernized and manufacturing moved or shrank (same source).
Machines kept moving the line
The two collars were never fixed groups. Each wave of technology moved tasks from one side to the other or removed them altogether.
- Mechanization took over much of the heaviest physical work in farming and factories, while creating jobs operating and maintaining machines.
- Office machines and computers took over routine calculation, filing and copying, and turned many clerical jobs into jobs working at a screen.
- Industrial automation and robotics moved factory workers toward supervising and maintaining automated lines.
The outcome was not always fewer jobs. The International Labour Organization points to bank tellers after cash machines spread in the 1990s: ATMs automated routine withdrawals, lowered costs and let banks open more branches, which raised demand for tellers in roles built around more complex customer service and advice (ILO Working Paper 140, 2025, citing Bessen, 2016).
What is different this time
Earlier waves mostly hit physical or highly routine tasks. Generative AI reaches into cognitive work: writing, summarizing, coding, analysis. The ILO notes that exposure to generative AI is highest in occupations with the most cognitive tasks and declines as tasks become more physical. It also notes that since 2023 these models have gained some agentic abilities: carrying out multi-step tasks on their own, working inside software environments and making decisions based on context (ILO Working Paper 140).
A 2023 study of the US workforce estimated that about 80% of workers could have at least 10% of their tasks affected by large language models, and about 19% could see at least half of their tasks affected (Eloundou et al., GPTs are GPTs).
So the line is moving inside the office this time. Many white collar tasks are becoming tasks that an AI system drafts and a person directs, checks and signs off.
Digital collar: a label for directing the machine
We use digital collar for work where the main job is setting tasks for AI agents and automation, checking what they produce, and owning the result. Unlike the older terms, it is not about clothing, and it does not describe a separate group of people. A warehouse supervisor running automated pickers, an accountant reviewing AI-prepared reconciliations and a developer reviewing agent-written code can all be doing digital collar work.
The pattern is the same one the history shows: tasks move, and the people who understand the work move toward deciding how it gets done. The difference is the speed and the reach of the change.
Where the collars stand now
Employers surveyed for the World Economic Forum expect structural change between 2025 and 2030 to create jobs equal to 14% of today's employment and displace jobs equal to 8%, a net gain of 78 million jobs. They expect the largest growth in absolute numbers in frontline roles such as farmworkers, delivery drivers and construction workers, and the largest declines in clerical and secretarial roles (WEF Future of Jobs Report 2025).
In other words, both of the old collars are still here. What is changing fastest is what happens at the desk.
Questions and answers
Where does the term blue collar come from?
From the sturdy blue work shirts, often denim or chambray, that manual workers wore because the color hid dirt and grease. Dictionary.com dates its first recorded use to 1945 to 1950.
Where does the term white collar come from?
From the white dress shirts and detachable collars worn by office workers, which came to stand for clerical, administrative and management work. It is the older term: Dictionary.com dates its first recorded use to 1920 to 1925.
Why is it called white collar?
It refers to the white dress shirts or detachable collars that male office workers wore through much of the nineteenth and twentieth centuries.
What is digital collar work?
Work where the main job is directing AI agents and automation, checking what they produce and owning the result. It is a kind of work, not a separate group of people.
Did earlier automation always cut jobs?
No. The ILO cites bank tellers: ATMs automated routine withdrawals, banks opened more branches, and demand rose for tellers doing more complex customer service and advice.