Knowledge Commons
HomeTopicsSeriesGuidesPopularAboutContact

The Poverty of the White Collar

Diploma, Language, Expertise and Insecurity: The Proletarianisation of Office Labour in the Age of Artificial Intelligence

59 min readAuthor: Oğuz Demirkapı
The Poverty of the White Collar

Dear young comrades,

In 1846 Proudhon published a book called The Philosophy of Poverty. Marx answered him a year later with The Poverty of Philosophy. Proudhon thought we could sift out the "bad sides" of capitalism and keep its "good sides". Marx showed that the good and the bad are two faces of the same relation, and that the machine produces both wealth and poverty with the same movement.

The title of this essay is a greeting to that debate. Because today, in the plazas, in the offices, at the screens, a new form of poverty is growing, and this poverty comes not from lack, but from the very middle of abundance. A section that has a diploma, that knows two languages, that writes code, that prepares reports, that makes presentations, that is called "professional", sees that its expertise does not mean security any longer. This poverty is not always an empty pocket. Sometimes it is the closing of the door, sometimes the handing of the contract to a subcontractor, sometimes the shift's stretching as far as the evening meal, and sometimes knowing that every email you wrote is read.

The word Marx uses in his German for "poverty" is Elend. This word tells not only of destitution, but also of homelessness, of estrangement, of the human being's losing control over his own life. The poverty of the white collar is precisely this: losing control over the product of one's labour, and equally over one's own process of work, one's time and one's future.

The occasion was an interview published on Bloomberg HT on 10 October. Fatih Cömert, Turkey General Manager of the international recruitment company Michael Page, said that, comparing 2026 with the previous year, they saw a fall of 32 per cent in white-collar recruitment (Bloomberg HT). Within a few days the news circulated under the headline "artificial intelligence reduced white-collar employment by 32 per cent" (Gazete Oksijen). Today, on 27 October, TÜİK announced the labour-force figures for September, and unemployment rose in one month from 7.8 per cent to 8.4 per cent.

In this essay, setting out from these two figures, we shall tell the condition of the white collar in five acts: the figures, the causes, the five faces of poverty, theory, and organisation.

First the summary, then the detail.

The figures in this essay have been compiled from publicly available sources as of 27 October 2026. The sources are linked both in the text and at the end. Documents to which a link could not be given are named.


First the Summary: What Does This Essay Say?

For the one in a hurry, the whole of the essay in twelve points:

  • 32 per cent is one company's observation, not the country's statistic. The figure is the contraction in Michael Page's own recruitment portfolio; mainly the middle and upper levels of corporate companies. It does not mean that white-collar employment across the country has fallen by 32 per cent. But it should be taken seriously as a sign.
  • The official picture is also darkening. According to the September figure TÜİK announced today, unemployment rose to 8.4 per cent, the number of unemployed to approximately 2,967,000, and youth unemployment to 14.8 per cent. Labour underutilisation is 31.2 per cent. By DİSK-AR's calculation, the number of the broadly defined unemployed is over 12 million. At ages 15–34, 6.8 million young people are neither in education nor in employment; among young university graduates this rate approaches one third.
  • The cause is not artificial intelligence alone. In the whole of the interview there are also the pressure of costs, the scissors of the exchange rate and inflation, the contraction of exporters, and the regional war. The news headline says only "artificial intelligence". To present the bill of the crisis as if it were the fate of technology makes the responsible party invisible.
  • But the effect of artificial intelligence is real, and its direction is clear. In the United States, the employment of workers aged 22–25 in the occupations most exposed to artificial intelligence fell by about 16 per cent relative to the others. In the United Kingdom, graduate advertisements are at their lowest level since 2020. In the United States, in the first seven months of the year, the ground of 112 thousand dismissals was stated as "artificial intelligence". According to the ILO, the occupational group most exposed is office work, and in these jobs women are far more numerous.
  • The first face of poverty is the door. Artificial intelligence, more than dismissing the worker, makes it easier not to hire. The lower rungs on which the occupation is learned are being dismantled; and the young graduate looks for work by passing through the sieve of the technology that is taking his place.
  • Its second face is security. At DeFacto Teknoloji, in September, four computing workers were dismissed. The union claims that subcontracted workers were taken into the positions that were emptied. The division of those who do the same work among different employers makes common bargaining difficult.
  • Its third face is the wage. The floor wage that TMMOB proposes for 2026 for the engineer, the architect and the city planner is a gross 94,500 TL; it is not binding. The minimum wage is a gross 33,030 TL. What determines the distance between the "value" of expertise and the actual return to labour is not the diploma, but bargaining power.
  • Its fourth face is time. Remote and hybrid work brought autonomy and a saving of time, but it erased the limits of the shift. The work of the worker who does not go to the office came to his home.
  • Its fifth face is surveillance. The Personal Data Protection Board ruled in September that the employer's technical access to corporate email did not give a right of unlimited inspection of content. The decision is important, but it discusses not the existence of surveillance, but its limit.
  • Nor is there a tool that would get the door opened. In the private sector, coverage of the collective agreement is 4.2 per cent, and unionisation among the young is 6.3 per cent. In the branch of Commerce and Office, the centre of gravity of office labour, only 6.6 per cent of 4.5 million workers are represented.
  • But the search has begun. The founding procedures of Bilişim-Sen, the computing workers' own union, were completed in November 2025. Even in civil-society organisations, workers are demanding a collective agreement. In the world, from Ireland to Spain, from the screenwriters in the United States to the white-collar union in Sweden, artificial-intelligence clauses are being written into collective agreements.
  • The class's stance: The white collar is now in a place where it cannot say "we are different". Our demand is not the banning of artificial intelligence, but the sharing of the time it creates: the shortening of working time without a cut in the wage, notification and union oversight in dismissals arising from artificial intelligence, a public obligation for entry-level employment, a limit on surveillance, and the removal of the obstacles in front of white-collar unionisation.

Now let us pass to the detail.


First Act: The Figures

One Interview, One Figure

The sentence that remained in the headline is a single figure. The whole of the interview, however, draws a much more complicated picture (Bloomberg HT, 10 October 2026):

SubjectWhat Cömert said
The general pictureComparing 2026 with 2025, a fall of 32 per cent in the white collar. Turkey is among the countries in difficulty, such as France and Germany.
ConsultancyThe need for consultancy firms diminished still further. Michael Page for this reason turned towards middle and senior roles.
Growing fieldsThe defence industry, data analytics, cyber security. In senior-executive searches, defence, energy and health.
Contracting fieldsClassic office roles. As sectors, the exporters most of all: automotive, the supplier industry, textiles.
The economic causeThat inflation and the exchange rate do not rise together made cost control difficult, and growth slowed.
The Middle EastWith the war, Turkish workers in Dubai and in the region want to return.
Artificial intelligenceThe rate of use rose from 24 per cent to 41 per cent, and this year to 60 per cent. 74 per cent of those looking for work are using artificial intelligence. Employers are doing the preliminary screening and the first interview with artificial intelligence, and this is more at entry level. Artificial intelligence will affect the white collar more than the blue collar.
The wageThe worker's first priority is the wage. On the global average 51 per cent, in Turkey 61 per cent.
The expectation2027 will to a large extent resemble 2026; recovery only in the second half of 2027.

Let us attend to two things.

The source of the figure. Michael Page is a recruitment and executive-search company; the product it sells is intermediation. The 32 per cent it sees is the contraction of demand in its own client portfolio. In the third-quarter results which the company's parent, PageGroup, announced on 13 October, gross profit fell by 5 per cent in the Europe, Middle East and Africa region, in which Turkey also sits, and by 5.5 per cent in the United Kingdom (AJ Bell). So the statement "Turkey, France and Germany are in difficulty" is at the same time a summary of the company's own regional balance sheet. This does not make what was said false; but while reading the 32 per cent we should know that we are also reading the contraction in an intermediary company's own market.

The causes. In the interview artificial intelligence is one cause, but not the sole cause. Into the news headline, only artificial intelligence came out. We shall return to its ideological function in the second act.

One more class note: the contraction of recruitment is not only the problem of the unemployed. When recruitment stops, the worker's possibility of changing jobs also narrows; when the possibility of changing jobs narrows, the power of bargaining over the wage also weakens. The employment crisis is also the crisis of the white-collar worker who is still at work.


The Official Picture: October's Latest Figure

Let us set the company's observation beside the official figure. According to the September 2026 labour-force statistics which TÜİK announced today, on 27 October:

  • The unemployment rate is 8.4 per cent; a month earlier it was 7.8 per cent.
  • The number of unemployed is approximately 2,967,000; an increase of approximately 227,000 in one month.
  • Youth (15–24) unemployment is 14.8 per cent; a month earlier it was 13 per cent.
  • Unemployment among women is 11.3 per cent.
  • The rate of labour underutilisation is 31.2 per cent.

These hold for the whole labour force; they do not show the white collar alone. But the direction is clear: the picture presented in August as "unemployment has receded" turned the other way one month later.

In TÜİK: Two Figures, One Country: 8.1% and 30.6% we had said "the gap between the narrow definition and the broad definition". This gap is still open. When fewer people say "I am looking for work", the official rate falls; but those who do half a job, and those who have lost hope and stopped looking, accumulate in the underutilised labour.

One table from five sources

So as not to lean on a single source, let us look at the same picture from different windows:

SourcePeriodWhat does it say?
TÜİK (official)September 2026Narrow unemployment 8.4%; youth unemployment 14.8%; women's unemployment 11.3%; labour underutilisation 31.2%.
TÜİK (NEET, compilation)2026, first quarterAt ages 15–34, 28 in every 100 young people are neither in education nor in employment: approximately 6.8 million young people. Among women the rate is 40%. Approximately one third of university graduates aged 15–24 (369 thousand out of 1.1 million) are neither in education nor in employment (Ekonomi Gazetesi, Alaattin Aktaş).
İŞKUR (official)August 2026Registered job seekers 2.34 million, up 1.6% in one year; a quarter of them are aged 15–24. The number of open jobs is 201,790, up only 2.7%. In the ten occupations given the most open jobs there is not a single office occupation (İŞKUR Monthly Bulletin).
BETAM (Bahçeşehir University)2026, second quarterEmployment stood still in one year (32.48 million), the labour force shrank by 233 thousand. The fall in the unemployment rate was not because jobs were created, but because people withdrew from the labour force. Employment in industry fell by 3.1% (Bloomberg HT).
DİSK-AR (union research)July–August 2026In July, broadly defined unemployment 30.6%, that is 12.3 million people; among women 40%. 5 million who want to work and have stopped looking, 4.5 million people in underemployment (DİSK-AR). According to the August figure the rate passed 31%, and the real number of unemployed rose by 322 thousand in one month (Evrensel).

The contraction Michael Page sees appears late in the narrow definition of the official statistic, and appears plainly in the broad definition. The white-collar unemployed person is most of the time not counted as "unemployed". Either he is put in "underemployment" by an internship and by project work, or after a while he stops looking and passes into the "potential labour force".

Let us also note a gap: in Turkey there is no official data set that measures the effect of artificial intelligence on employment by occupation and by age. So long as the SGK's payroll data, broken down by occupation and by age, is not opened to researchers, the discussion will remain condemned to the portfolio observations of recruitment companies. Access to data is also a class matter: the employer sees his own data; the worker does not.

Graduates: those waiting at the door

According to TÜİK's Higher Education Employment Indicators, at least one fifth of the bachelor's graduates who graduated between 2015 and 2024 are not in registered employment in 2025 (eleman.net compilation):

FieldTime to first entry into a jobNot in registered employment / working outside the field
Health and welfare9 monthsAmong physicians, about one fifth outside the field
Engineering, manufacturing, construction11.6 monthsAbout one third outside the field
Social sciences, journalism, information17.7 months31.3% outside employment; 80% of those employed are outside the field
Business, administration, law18.3 months-

The longest waiting time is among graduates of business, administration and law. These are the graduates of the classic white collar, that is, of the report, the contract, the presentation, the accounts. The field of which Michael Page says "classic office roles are contracting" overlaps with the field in which graduates wait longest at the door.


Second Act: The Causes

The Exchange Rate, the Crisis, and the Language of Fate

The news headline said "artificial intelligence". The interview itself said the exchange rate, inflation and exports. These two have to be separated.

The reason exporters were squeezed in Turkey in 2026 is that inflation rose faster than the exchange rate, that is, that the Turkish lira gained value in real terms. This is not a natural event, but a conscious choice of monetary policy: in order to bring inflation down, the exchange rate is held under pressure. The price is paid in export industry and in the white-collar staff around it. Automotive, the supplier industry and textiles, the fields Cömert counts as the sectors most affected, are precisely the first addressees of this policy.

Let us tell the same contraction in two sentences:

  • "Monetary policy squeezed the exporter, companies are cutting costs, white-collar recruitment has stopped."
  • "Artificial intelligence has come, white-collar jobs are diminishing."

The first sentence has a subject: the state that makes the policy, and the company that cuts the cost onto the worker. This subject can be criticised, and a demand can be organised against it. In the second sentence there is no subject, there is only technology. A strike is not made against technology, and a collective agreement is not imposed on technology. To present the bill of the crisis as if it were the fate of technology makes the responsible party invisible. For this reason the headline "artificial intelligence is eating jobs" serves, above all, the employer who freezes recruitment in order to protect the profit margin.

This is not to deny the effect of artificial intelligence. The two processes feed each other. Periods of crisis are the periods in which companies accelerate the automation they had already planned. When the pressure of costs comes, the decision "let us not fill this position again, let us distribute the work to those who remain, with artificial-intelligence tools" becomes easy. And when the crisis passes, that position does not come back. As Marx tells it in Capital, the machine spreads fastest in periods of depression, because it is then that capital most looks for the way to economise on living labour.

The effect of artificial intelligence is determined by the companies' choice as much as by the capacity of the technology. When, in a software team, artificial intelligence speeds up the production of code, four results are possible: more work with the same staff, a shorter delivery time, turning towards new products, or fewer personnel. It is not the algorithm that decides which is chosen, but management. And the only power that can change management's decision is the organised labour opposite it.


Canaries: What Does the Effect of Artificial Intelligence Say?

The soundest data on the effect of artificial intelligence on employment comes from the United States. Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, of the Stanford Digital Economy Laboratory, examined millions of records of the largest payroll company in the United States, and gave their findings the name "Canaries in the Coal Mine" (Stanford Digital Economy Lab).

The core of the finding: in the occupations most exposed to artificial intelligence (such as software development, customer service and accounting), the employment of workers aged 22–25 fell in a marked way relative to the other occupations. Among older and more experienced workers in the same occupations there is no such fall. In the February 2026 update the researchers also tried to separate out the effect of interest rates: even with the tightest control, the relative fall of young workers in the most exposed occupations rose to approximately 16 per cent up to October 2025, and there is no turning back (Stanford, February 2026). The researchers also write explicitly that they do not count artificial intelligence as the sole cause.

The pattern is this: artificial intelligence, more than dismissing the worker, makes it easier not to hire. The senior worker stays in place; artificial intelligence turns, in his hand, into a tool of productivity. The machine does the work of the one who was going to come. Without news of a dismissal, silently, a generation remains outside the door.


The World's Mirror

Let us set the picture in Turkey beside the world. A warning: countries measure unemployment with different age ranges, periods and definitions. The figures are not for a one-to-one comparison, but to give an idea of direction and of magnitude.

General and youth unemployment
Country / regionGeneral unemploymentYouth unemployment (under 25)Period and source
Turkey8.4%14.8%September 2026, TÜİK
Euro area6.4%14.9%July 2026, Eurostat
European Union6.1%15.1%July 2026, Eurostat
Germany4.0%7.2%July 2026, Eurostat
France8.3%20.7%July 2026, Eurostat
Spain-22.9%July 2026, Eurostat
Sweden-24.4%July 2026, Eurostat
OECD average5.0%11.5%February 2026, OECD
United States4.3%-March 2026, OECD compilation

Sources: Eurostat, OECD compilation (DevelopmentAid).

It is easy to say "youth unemployment in Turkey is the same as the European average", but it is misleading. The unemployment rate counts only those who are looking for work. In Turkey a large part of young people are not even looking for work: about 24 in every 100 young people aged 15–24, 28 aged 15–34, and 40 among young women, are neither in education nor in employment. While in Europe the young unemployed person is counted as "unemployed", in Turkey a considerable part of the young unemployed turns into a mass that sits at home and does not appear as unemployed in any statistic. The rate hides a generation that has been pushed outside the statistic.

The difference between Germany and France is also instructive. Their general unemployment is not very far apart, but in youth unemployment Germany is 7.2% and France 20.7%. One of the explanations most often put forward for this difference is the dual vocational-training system in Germany: young people work at the workplace, simultaneously with school, as paid apprentices, and the rung of entry into the occupation is a legal and institutional obligation. The system has problems of its own; it is not a model on its own. But it says this: when the bottom rung of the ladder is not left to the market, young people do not remain outside the door.

New graduates and the white collar
IndicatorFigurePeriod and source
Turkey: those neither in education nor in employment among university graduates aged 15–24~33%2026, first quarter, TÜİK
United States: unemployment of new graduates aged 22–275.6% (7.2% among peers without a diploma)2026, first quarter, New York Fed
United States: relative employment loss of those aged 22–25 in the occupations most exposed to artificial intelligence~16%Up to October 2025, Stanford
United States: entry-level job advertisements~35% decreaselast 18 months (May 2026), Revelio Labs
United States: graduates of 2023 who entered a job that does not require a diploma52%Burning Glass Institute
United Kingdom: graduate job advertisementsannual fall of 7%, the lowest since 2020June 2026, Indeed
United Kingdom: share of job advertisements that mention artificial intelligence9.4% (a record)June 2026, Indeed

Sources: New York Fed (Bloomberg Law), Revelio Labs (GZT), Burning Glass (eWeek), Indeed (Workplace Journal).

The pattern is the same from country to country, with a difference of scale: graduate advertisements are diminishing, advertisements that want artificial intelligence are increasing. Capital on the one hand narrows recruitment, and on the other expects the remaining worker to get more work done with artificial intelligence.

The "artificial intelligence" ground in dismissals
IndicatorFigurePeriod and source
Total announced dismissals in the United States477 thousand (41% less than the same period of 2025)January–July 2026, Challenger
Of these, those whose ground is "artificial intelligence"112,713 (24% of the total)January–July 2026, Challenger
Dismissals on the ground of artificial intelligence in the whole of 202554,8362025, Challenger
Consecutive months in which artificial intelligence was the ground most often given5March–July 2026, Challenger

Source: Challenger, Gray & Christmas, July 2026, May 2026 compilation.

In the United States, total dismissals have diminished by 41% compared with last year, but those announced as "because of artificial intelligence" have, in seven months, risen to twice the whole of 2025. There are two readings, and both may be correct. The first is that artificial intelligence is really becoming the instrument of the liquidation of jobs. The second is that artificial intelligence is becoming the most usable ground of dismissal. To say "we are cutting costs" looks bad to the shareholder, and provocative to the worker. To say "we are carrying out an artificial-intelligence transformation", however, looks to the shareholder like a promise of the future, and to the worker like fate.

Whose job is more exposed?

According to the global index the ILO published in May 2025 (ILO):

IndicatorFigure
Share of jobs in the world potentially exposed to generative artificial intelligence25%
This share in high-income countries34%
Occupational group at the highest exposureOffice and secretarial work
Share, in women's employment, of the highest-risk jobs in high-income countries9.6%
Share of the same jobs in men's employment3.5%

The field most affected is the classic office work of which Michael Page says "it is contracting". In these jobs women are far more numerous. Recall that in Turkey women's unemployment is 11.3%, and that the rate of being neither in education nor in employment among young women is 40%. The debate on artificial intelligence cannot be conducted blind to gender.


Third Act: The Five Faces of Poverty

The figures draw a frame. Now let us look inside the frame. The poverty of the white collar is not lived in a single form; it has five faces.

The First Face: The Door

Let us think of how an occupation is learned in an office. A newly graduated lawyer first writes a draft petition, and searches the case law. A newly graduated economist first cleans a table, and summarises a report. A newly graduated programmer first writes tests, and corrects small errors. A newly graduated marketer first compiles a competitor analysis. These jobs are tedious, but they are the school of the occupation. The senior worker became senior by passing through these rungs.

Today language models do all of these jobs in seconds, at an acceptable quality. The senior lawyer has the machine write the draft and corrects it himself; there is no longer a need for the junior lawyer. But when the senior retires ten years later, the junior who would take his place has never been trained.

This is the classic contradiction between the short-term rationality of capital and the long-term reproduction of society. From the point of view of each single company, not hiring a junior is rational: it pays the cost of the training, and the worker who passes to another company can carry off the return. But when all the companies give the same decision, the occupation cannot reproduce itself. Capital does not resolve this contradiction of itself. Entry-level employment has to be a matter not of the market, but of public regulation.

There is also an irony. According to Cömert, employers are doing the preliminary screening and the first interview with artificial intelligence, and this is applied most of all at entry level. On the other hand, 74 per cent of those looking for work are also preparing their application with artificial intelligence. The young graduate tries to pass, through the sieve of that same technology, the curriculum vitae he prepared with the technology that is taking his place. Two machines are talking to each other; the human being in between is waiting to be screened out.

The Second Face: Security

For those who can get in through the door, the ground is not firm either. An event in the last week of September made visible the debate on the job security of computing labour.

According to the Evrensel report dated 23 September, computing workers were dismissed at DeFacto Teknoloji (Evrensel). The report conveyed that three workers were dismissed, and that one worker left after completing his premium days; in other reports the event passed as the dismissal of four computing workers. The Computing Workers' Union (Bilişim-Sen), arguing that the problems in the company's inventory management were loaded onto the workers, demanded reinstatement. The union also claimed that subcontracted workers were taken into the positions that had been emptied. These claims have to be assessed together with the company's grounds and with independent legal findings; it would not be right to give a final judgement before the lawsuit or the mediation has concluded.

But the class meaning of the event is clear, independently of its details. The sentence in the headline of the report sums this up: "The workers will not pay the price of growth." In a company that is growing, widening its shop network and investing in digitalisation, even the computing worker who produces the technology itself can be seen as a cost item that is easily replaced.

Outsourcing does not in itself mean insecurity. But the division of workers who do the same work among different employers, different wage systems and different contracts makes common bargaining difficult. If three of five people sitting in the same team are on the payroll of different companies, who will bargain with whom, and opposite whom? Subcontracting, in the office as in the factory, serves this above all: dividing a collective subject into individuals and into companies.

The Third Face: The Wage

The white collar is known as a section that "earns well". Let us look at the figures.

The monthly gross floor wage that TMMOB proposes for 2026 for the engineer, the architect and the city planner is 94,500 TL. This is a proposal; it is not a binding legal minimum wage. In the same year the legal gross minimum wage is 33,030 TL. That is, the occupational organisation says that an engineer's labour ought to be about three times the minimum wage; but no employer has to comply with this proposal.

The issue here is not only the difference between the two figures. What determines the distance between the "market value" of expertise and the actual return to labour is not the diploma, but bargaining power. The recommended wage, so long as it is not binding, consists only of a wish. The only tool that can make it binding is the collective agreement. The engineer, the architect, the programmer can learn from a professional chamber what his wage ought to be; but in order to be able to receive that wage he has to be organised.

To read the white-collar wage correctly, seniority, expertise, dependence on the employer, the cost of housing, overtime, and the course of the wage against inflation have to be looked at together. The "high" salary of a programmer in Istanbul who pays rent, who works fifty hours a week, and whose wage is updated once a year, melts a little more each month beside inflation. Cömert's own data confirms this too: in Turkey 61 per cent of workers count the wage as their first priority, and the global average is 51 per cent.

A warning is also necessary: it would not be right to place all white-collar workers at a single income level or in a single class position. The positions of the senior executive, of the freelance professional working on his own account, and of the waged specialist are different. The dependence on the employer of a software engineer who receives a higher salary does not disappear; but he is not in the same position as a general manager either. We shall open this distinction in the fourth act.

The Fourth Face: Time

Remote and hybrid work, which spread with the pandemic, was presented to the white collar as a promise of freedom: no traffic, no office, you set your own hours yourself. A part of the promise came about. But the cost also emerged.

The study by Hümeyra Taşçıoğlu, published on 20 October 2026 in the journal İmgelem, examines the experience of 19 white-collar workers in the computing sector in Turkey. According to the study, remote and hybrid work provides a saving of time and autonomy; but the limits of the shift grow indefinite, an expectation of being constantly reachable forms, and the workers feel anxiety about career possibilities. This is not a quantitative study representing the sector; but it points to an important contradiction: even if the worker does not go to the office, the demands of the work extend as far as his home.

Marx tells the struggle over the length of the working day in one of the longest chapters of Capital. The eight-hour working day, as we showed in What Did the October Revolution Give the World?, was the gain of a class struggle before it was a clause of a law. Remote work silently wears this gain away. In the factory, at the end of the shift, the whistle blows; at home there is no whistle. The Slack notification comes at eleven at night as well. The shift begins to be measured not by the hour, but by availability.

According to TÜİK, average actual weekly working time rose in August to 42.5 hours. That working time does not shorten while artificial intelligence increases productivity, and on the contrary lengthens, confirms the main thesis of this essay once more: the increase in productivity turns not into the worker's working less, but into capital's appropriating more.

The Fifth Face: Surveillance

The white collar's work increasingly leaves a digital trace: email, corporate messaging, access records, the history of the code repository, meeting recordings, screen time. These traces are gathered for information security; but the same traces can also be used to oversee the worker.

An important legal development of 2026 took place in this field. The Personal Data Protection Board, by Principle Decision no. 2026/2035 dated 16 September 2026, set a limit on employers' oversight of corporate email and of other work channels of communication. According to the decision, the employer's having a technical authority of access to a system does not mean that he has a right of unlimited access to the content of the workers' communication. The oversight has to be clearly notified in advance, limited to a legitimate purpose and proportionate, and, where necessary, less intrusive methods have to be preferred. In place of a general and continuous inspection of content, oversight based on a concrete ground is foreseen.

The decision is important, and it should be used. But let us also see its limit: the decision does not prohibit workplace surveillance; it brings to it a legal limit and an obligation of justification. Gathering data in order to ensure the security of the system, and surveilling the worker's behaviour without limit, are not the same thing; but what will separate these two is not the text of the decision, but a workers' organisation that can oversee this distinction in the workplace.

In the chipped-tracking debate at Astor, and in Anthropic's Activist Monitoring System, we had shown how surveillance extends from the factory to the company headquarters. In the plaza, surveillance works not by chip and camera, but by log and by algorithm; but the logic is the same. And this logic has now come as far as the door of recruitment: preliminary screening done with artificial intelligence means the candidate's entering surveillance before he is yet a worker.

Surveillance has one more consequence for organising. The corporate channels the employer can see also make visible the correspondence of the workers among themselves. The first step of organising in the plaza, that is, saying "let us talk about this with the others too", can for this reason be riskier than in the factory. The KVKK decision is a tool for reducing this risk; but by itself it is not sufficient.


Fourth Act: Theory

A Marxist Reading: Organic Composition, the Reserve Army, Proletarianisation

Organic composition. Marx divides capital in two: constant capital, invested in the machine, the building and the raw material, and variable capital, invested in the wage. The long-term tendency of capitalism is the growth of the first relative to the second: fewer workers, more machinery, for the same output. In the 19th century this was lived at the weaving loom, in the 20th century on the assembly line. In the 21st century it came to the office. The artificial-intelligence model is the weaving loom of the office.

Crystallised labour. This loom has a new side. Language models were trained on texts written by millions of people, that is, on the common mental labour of society. What we called the expropriation of the general intellect, in The Expropriation of the General Intellect and the AGI Illusion, is this: the reports, the emails, the contracts and the code that white-collar workers have written for decades became the raw material of the machine that takes their place. The worker is being substituted by a machine produced from his own labour. What we called mental Taylorism, in The Expropriation of the General Intellect and the AGI Illusion, is the other face of this: office work, just like factory work, is fragmented, measured, standardised, and in the end handed over to the machine.

The industrial reserve army. Every young graduate who is not hired joins the mass Marx called the industrial reserve army. The function of this army is to pull down the wage of those who have a job: "if you don't want it, there are many waiting at the door." The passing of underutilised labour beyond 31 per cent, and graduates' waiting 18 months to enter a first job, is the measure of this pressure.

Relative surplus-value. The work of the junior who is not hired does not disappear; it is distributed to the one who remains. The senior worker, with artificial-intelligence tools, does the work that three people used to do. His wage stays the same, his load increases. Marx calls this relative surplus-value. When we wrote on Anthropic's survey of 81,000 people, A Class Analysis of the World of Artificial Intelligence in Light of the Anthropic Survey, we had called this the "treadmill paradox": you run faster, but you are in the same place.

Is the White Collar a Single Class?

No. "White collar" is not a class, but a category of appearance. The general manager of a holding, and the young woman who works in the call centre of the same holding, are both white-collar. The first carries out the function of capital; the second is a wage labourer. The broad section in between, that is, the middle-level manager, the specialist, the engineer, the analyst, was for a long time a subject of debate in the Marxist literature.

Harry Braverman, in Labor and Monopoly Capital, which he wrote in 1974, showed that office work, just like factory work, was fragmented, standardised and deskilled. Typing pools, accounting machines and punched-card operators were the artificial intelligence of his time. Erik Olin Wright defined the sections of managers and specialists as "contradictory class locations": although they are wage labourers, they carry in the workplace an authority of oversight and a privilege of expertise. One foot is in labour, one foot is in the oversight function of capital.

What artificial intelligence is doing today is wearing away the "privilege" side of this contradiction. Expertise is transferred to the machine, oversight is handed to the algorithm. What remains is the bare relation of wage labour. The contradictory location is being resolved, and the direction in which it is resolved is clear. This does not mean that all white-collar workers are being impoverished at the same speed; nor does it mean that all occupations face the same threat. But it shows that the direction is the loosening of the bond between expertise and security.

The idea that "the white collar is the middle class" had three supports: job security, the career ladder, and the holding of mental labour apart from manual labour. All three are coming apart. Security by the flexible contract and the subcontractor, the ladder by the dismantling of the bottom rung, the privilege of mental labour by artificial intelligence. The white collar is beginning to live what the blue collar has been living for decades. This can be told as a story of decline, but from the standpoint of class it is a story of clarification.

What Do the Growing Fields Say?

Cömert's list of growing fields is short, but it says a great deal: data analytics, cyber security, the defence industry.

The first two are the infrastructure of the new regime of accumulation of monopoly capitalism: data analytics gathers and processes the data, cyber security protects that data. As we showed in Bubble or Debt? Meta's 145 Billion Dollars and the Class Character of the AI Bubble, hundreds of billions of dollars are flowing into this field, and this capital can draw its profit only by substituting living labour. That is, a part of the growing fields is producing the very technology that contracts the contracting fields.

The third, the defence industry, is a demand not of the market but of the state. In a period when war has surrounded our region, the directing of qualified white-collar labour to the production of weapons is a choice. The same engineer, the same programmer, the same data analyst could also have worked for earthquake safety, public health, climate adaptation, or public transport. It is not technology that decides which field will be a "growing" one, but the politics that makes the budget and directs capital.

In short, white-collar labour is being reorganised, but not around social need, around surveillance, data and the weapon.


Fifth Act: Organisation

One reason this picture is lived more heavily in Turkey is the white collar's lack of organisation. The wave of artificial intelligence strikes a section that has no collective tool. In Germany there is a system that protects the apprentice, in Sweden a union that puts working time on the bargaining table, in Ireland an organisation that signed an artificial-intelligence clause with a bank. In Turkey, what the plaza worker has opposite him is, most of the time, only the human-resources department.

The Absence of Organisation, in Figures

According to the July 2026 statistics of the Ministry of Labour and Social Security, about 2.43 million of 17.63 million workers are union members; the official rate is 13.8 per cent. Of the 246 workers' unions in activity, only 60 can pass the 1 per cent branch threshold required for the authority to make a collective agreement. DİSK-AR's July 2026 Report on Trade Union Rights deepens this picture (bianet):

  • Of 18.7 million workers, including those in unregistered work, only 2.4 million are in a union: 87 in every 100 workers are not in a union.
  • Unionisation in the private sector is 5.7 per cent, and coverage of the collective labour agreement is 4.2 per cent. In the private sector, the wage of only 4 in every 100 workers is set by collective bargaining.
  • In general coverage of the collective agreement, Turkey, at 9.7 per cent, is second from last among the 27 countries compared, after Mexico.
  • Unionisation in the 15–24 age group is 6.3 per cent; in all of the 21 branches, young people are below the branch average.

These figures do not directly show the unionisation of the private-sector white collar, but they show the general structure. And they say one thing plainly: the section artificial intelligence strikes most is the young; the young are already the least organised section. The generation waiting outside the door does not possess the tool that would get the door opened, either.

Let us look at the branches in which the white collar is dense. According to the assessment of the Fişek Institute on the July 2026 statistics, the representation rates of the unions that passed the threshold are as follows (Fişek Institute):

Branch of industryNumber of workersRepresentation rate of the unions that passed the threshold
09 - Banking, Finance and Insurance353 thousand30.4%
17 - Health and Social Services793 thousand30.3%
08 - Press, Publishing and Journalism92 thousand12.5%
07 - Communication209 thousand12.3%
10 - Commerce, Office, Education and Fine Arts4.5 million6.6%
Comparison: 20 - General Works1.1 million52.6%

The heart of the table is branch number 10. The greater part of the office worker, the sales representative, the call-centre worker, the private-school teacher, the specialist of the consultancy firm, and the computing worker (The Computing Worker's Handbook) is registered in this branch. 4.5 million workers, more than a quarter of the insured workers in Turkey. In this branch 31 unions are in activity, only 3 of them can pass the threshold, and the representation rate is 6.6 per cent. The centre of gravity of the white collar is one of Turkey's most crowded and most unorganised branches of industry.

(These rates are a "representation rate"; the members of unions below the threshold are not counted, and because of passive members the figure may also be somewhat inflated. But the order of magnitude does not change.)

Why Unorganised? Four Obstacles

The first obstacle: ideological. For long years the plaza worker defined himself not as a worker, but as a "professional". He counted the union as the blue collar's business, and the strike as the factory's subject. The language of human resources fed this perception: the worker is a "talent", the team is a "family", the wage is a "package", dismissal is "we are parting ways". This language shows the relation of wage labour as if it were a career journey. In a workplace where everyone bargains his wage in secret, and where talking about one's salary is counted as shameful, it is hard to build solidarity; because the first condition of solidarity is to see the common situation.

The second obstacle: legal. Law no. 6356 requires, for the authority to make a collective agreement, both a 1 per cent threshold in the branch and more than half of the workers at the workplace, and 40 per cent in the enterprise. In a plaza company where jobs are changed often, and where the workers are in different cities and work remotely, organising more than half at the same time is much harder than in a factory. Besides, a software company, an advertising agency and a call centre are gathered in the same broad branch; an administrative classification, rather than an occupational commonality, determines the frame of organising. In banking there is another obstacle: in 2016, by a state-of-emergency decree-law (Decree-Law no. 678), strikes in banking services were brought within the government's authority of "postponement" (Evrensel). Postponement, at the end of 60 days, puts the decision of the High Board of Arbitration in the place of the strike. Even in the white collar's most organised branch, the organisation's strongest tool can in effect be taken from its hand.

The third obstacle: the form in which the work is organised. White-collar work is increasingly scattered: hybrid and remote work, project-based contracts, freelance work, subcontracted consultancy, seniorities shorter than two years. As we saw in the DeFacto example, those who do the same work can be divided onto different payrolls. The place of the worker who came together at the change of shift in the factory has been taken by workers who see one another only on the screen. This scatteredness is not a coincidence; it is capital's choice to the extent that it makes collective bargaining difficult.

The fourth obstacle: employer practice. Article 25/II of the Labour Law gives the possibility of dismissal without compensation on the ground of "situations that do not conform to the rules of morality and good faith"; DİSK has been saying for years that this article is a tool of arbitrary dismissal. In the white collar, performance evaluation is added to this: dismissal on account of union activity is forbidden, but the ground of "low performance" or of "restructuring" is always ready. And there is also the surveillance we described in the fifth face.

New Searches for Organisation

The picture is dark, but it is not empty. In the last months of 2026, two developments showed that the labour of computing and the wider labour of the white collar can no longer be taken up separately from one another.

The computing workers' own union. According to the activity report of the Computing Workers' Union (Bilişim-Sen) dated 13 June 2026, the union's official founding procedures were completed on 20 November 2025. In 2026 the Izmir unit was formed, a games-sector working group was set up, and organising meetings were held in Istanbul and in Izmir. In September it came before the public with a demand for reinstatement against the dismissals at DeFacto Teknoloji. These works do not yet show that a determining level of unionisation across the sector has been reached. But they are a new example of programmers, game developers and other computing workers organising around common occupational and economic problems. In The Computing Worker's Handbook, when we said "computing worker", we meant not only the one who writes the code, but everyone from the network administrator to the designer, from the product manager to the producer of content. This breadth is the general lesson of white-collar organising.

Even in civil society. The white collar's lack of organisation is lived not only in companies run for profit, but also in institutions aimed at a social purpose. The workers of the Community Volunteers Foundation (TOG) had organised in DİSK Sosyal-İş in 2010, and had signed six collective agreements up to 2021. In 2021 union activity ended; the workers organised again later. In May 2026, one day before the meeting of the union with the management that had been awaited for months, workers, among them project managers, were dismissed (bianet). The foundation gave contraction and the reduction of costs as the ground; the workers argued that the dismissals were connected with organising. In September the workers carried the demand for a collective agreement before the public again (Evrensel). The talks conducted with DİSK Sosyal-İş ended after four meetings; the workers want their existing rights secured, and the talks to begin again.

This example is unpleasant, but instructive. The employer–worker relation is a class relation, independently of the employer's intention. An institution that works for social peace, for youth and for volunteering can also, when it behaves as an employer, see the union as a cost and as a threat. Institutions that aim at the benefit not of capital but of society ought to be the first to recognise their own workers' right to organise. This is the test of their consistency with their own values.

The "white collars" of Gezi. It had been seen before, in Turkey, that the white collar can be a collective subject: in the Gezi resistance of 2013, plaza workers went into the street with a "white-collar" identity, and defined this identity for the first time as a political subject (bianet). That this energy could not be carried into the workplace is a part of the lack of organisation we are living today. The lesson is this: being visible in the street is not enough; one has to be organised in the workplace.

How Does the White Collar Organise in the World?

There is no law that the white collar cannot organise. In the world there are strong white-collar unions, and in recent years they have carried artificial intelligence directly to the bargaining table:

CountryOrganisationWhat did it do?
SwedenUnionen (the country's largest white-collar union)In September 2026 it gave priority, for the coming round of collective bargaining, to the demand for a real wage increase and for the shortening of working time once again (WageIndicator).
IrelandFinancial Services Union (about 10 thousand finance workers)In March 2025, an agreement with the Bank of Ireland: artificial intelligence will not be used in employment decisions, retraining will be given priority instead of dismissal, and in artificial-intelligence processes the decision will remain with the human being (Partnership on AI).
SpainNational collective agreements for banking, credit institutions and insuranceThe worker's right not to be subjected to a decision based only on automatic variables; the obligation to notify union representatives of the data that feeds the algorithm, of the logic of its operation, and of its results (IBA).
United StatesWriters Guild of America (screenwriters)At the end of the strike that lasted months in 2023: artificial-intelligence output will not be counted as "source material", the screenwriter will not be able to be forced to use artificial intelligence, the artificial-intelligence text that is used will be notified, and the union will be able to object to the use of members' texts in the training of models (Fortune).
United StatesNewsGuild-CWA (journalists) and ZeniMax workers attached to MicrosoftAccording to NewsGuild there is an artificial-intelligence clause in 85–90 collective agreements. In the ZeniMax agreement, artificial intelligence is to "support the worker, not replace him", and there is an obligation of notification to the union and of bargaining for new systems (Allwork).

The common lesson: the most effective protection against artificial intelligence is not the law, but the collective agreement. The law remains general, and comes late; the collective agreement is written according to the workplace's concrete technology and its concrete job description. The WGA example shows one thing more: the white collar too can strike, and can win. The screenwriters, by stopping work for months against the model's taking their place, a model trained on their own labour, had an artificial-intelligence clause written into a contract for the first time.

To see the scale, a comparison: even in the United States, known for its hostility to unions, the rate of workers within union coverage was 11.2 per cent in 2025. In Turkey, coverage of the collective agreement in the private sector is 4.2 per cent.

Where Can One Begin?

Putting artificial intelligence on the table of the existing organisations. In banking, finance and insurance the representation rate is above 30 per cent; in health it is so as well. In the press and publishing, and in communication, there is an organised base of around 12 per cent. That the unions in these branches demand an artificial-intelligence clause in their next collective agreement can be the first concrete step. The banking agreement in Spain, and the Bank of Ireland agreement in Ireland, are texts that can be taken as examples.

Organising in branch number 10. A branch in which 6.6 per cent of 4.5 million workers is represented is, at the same time, Turkey's largest potential for organising. Remaking the regulation of branches according to occupational commonality, defining a separate branch for computing, and removing the thresholds, removes the legal obstacles in front of this potential. The removal of the thresholds is also among DİSK's basic demands.

Forms of solidarity before the union. Even if the authority to make a collective agreement looks distant, there is no need to wait. Wage transparency (those who do the same work sharing their wages with one another), a map of the workload (a record of the positions that are not filled again, and of the work that is distributed), safe channels of communication, and solidarity networks based on the occupation, prepare the ground of union organisation.

Making the professional chambers part of the wage struggle. The floor wage TMMOB recommends is not binding; but it is a reference point. When the wage proposals of the professional chambers are joined with the collective-agreement demands of the unions, they can turn from a wish into a bargaining target. The division of labour between the chamber and the union is one of the least discussed, but most fruitful, fields of white-collar organising.

A common demand with the blue collar. The white collar's power of organising on its own is limited; but it can unite with the blue collar around common demands. The shortening of working time, the regulation of 25/II, the removal of the thresholds, and the ending of strike bans and of postponements, are the demand both of the factory and of the plaza. As we discussed in The Computing Worker's Handbook, the warehouse worker, the courier and the programmer are links of the same value chain. The DeFacto example shows this nakedly: in the same company's shop the cashier works, in its warehouse the warehouse worker, in its technology unit the programmer, and all of them pay the same "price of growth".

A Sample List of Demands for an Artificial-Intelligence Clause

Setting out from the examples in the world and from the debates in Turkey, we can compile as follows the clauses that could enter a collective agreement or a workplace protocol:

DemandContent
Notification and consultationWritten notification to the union or to the workers' representatives, and bargaining, before a new artificial-intelligence system is put into operation.
A commitment not to dismissNo dismissal on the ground of an artificial-intelligence application; priority of paid retraining and of internal placement for the positions affected.
A ban on subcontractingPositions on the permanent staff that are emptied on the ground of artificial intelligence or of restructuring are not to be filled by a subcontractor or by outsourcing.
An obligation to fill againEntry-level positions that are emptied are not to be closed on the ground of artificial intelligence.
A share of productivityThe increase in productivity arising from artificial intelligence is to be shared as a wage increase or as a shortening of working time.
The right to be offlineNo obligation to answer emails and messages outside the shift; a written definition of the limit of the shift in remote work.
A human decisionDecisions on recruitment, performance, promotion and dismissal are not to rest on the algorithm alone; a right of objection and of human assessment.
Algorithmic transparencyAn explanation to the representatives of which data the system used in decisions about workers uses, and of how it operates.
A limit on surveillanceThe application in the workplace of the Personal Data Protection Board's Principle Decision no. 2026/2035: inspection of content to be notified in advance, tied to a concrete ground, and union correspondence not to be monitored.
The property of the labourThe texts, the code and the documents the workers produce are not to be used in the training of models without consent.

This list is a beginning; every workplace should adapt it to its own technology and its own job description. But what matters is this: the only way to have these clauses written in is the existence of an organisation that will bring them to the table.


What the Headline Says, What the Class Must Ask
What the news saysWhat the class must ask
There is a fall of 32 per cent in the white collar.Whose portfolio is this figure? Why is underutilised labour above 31 per cent?
Artificial intelligence is reducing white-collar jobs.Why are exchange-rate policy and the cost cut not in the headline?
Companies are turning towards the middle-level and senior specialist.Where will companies that do not hire juniors find the specialist ten years later?
Employers are doing the interview with artificial intelligence.Does the screened-out candidate have a right of objection, and who oversees the algorithm?
The company is "restructuring".Who is being taken into the emptied post, on what contract, and at what wage?
The engineer's floor wage should be 94,500 TL.Where is the collective agreement that will make this wage binding?
Remote work provides flexibility.Who determines the hour at which the shift ends?
The employer keeps a log for the security of the system.Who oversees the line between security and surveillance?
Data, cyber security and defence are growing.Who decided that qualified labour be directed to the weapon and to surveillance?
2027 too will resemble 2026.Who will do what for the generation of graduates who will wait at the door for one year more?

So What Is to Be Done?
To watch: five headings of the period ahead
  1. Job security: the rights of workers in decisions of restructuring and of outsourcing. Follow the outcome of the DeFacto and TOG processes.
  2. The wage and working time: how the increase in productivity arising from artificial intelligence is reflected in wages, in the number of posts, and in the hours of the shift.
  3. Digital oversight: the transparency of the measurement of performance, the protection of personal data, and the auditability of algorithmic decisions. How the KVKK decision is applied in workplaces.
  4. Union organisation: the legal and institutional obstacles white-collar workers meet while using the right of collective bargaining.
  5. The position of new graduates: the effect of the use of artificial intelligence in entry-level jobs on occupational experience and on the rungs of a career.
To read
  • Read the figure with its source. The portfolio data of a recruitment company and the TÜİK data are not the same thing. Each month, set beside the unemployment rate the underutilised labour, young women's unemployment, and graduate employment.
  • Ask the cause of headlines that say "because of artificial intelligence". In every piece of news of a dismissal or of a freeze on recruitment, ask two questions side by side: how much is technology, how much is the cost cut? What happened to the company's profit, and to executive bonuses, in the same period?
  • Draw the map of your own workplace. Which positions were not filled again in the last two years? Which jobs were distributed, with artificial-intelligence tools, to those who remained? Who is on a subcontractor's payroll? How did your workload and your shift change? Sharing this information with a friend of yours is the first step of organising.
To demand
  • The shortening of working time. If artificial intelligence increases productivity, the return ought to be not unemployment, but less work for everyone. DİSK-AR's proposals show the way: the reduction of weekly working time to 37.5 hours without a loss of income, the lowering of the annual overtime limit from 270 hours to 90 hours, the regulation anew of Article 25/II, which opens a door to arbitrary dismissal, an end to the use of the Unemployment Insurance Fund outside its purpose, and the removal of the union thresholds (DİSK-AR). These demands hold for the white collar in just the same way; the section that does overtime in the plaza the most, under the name of "flexibility", is the white collar.
  • Oversight of dismissals arising from artificial intelligence. It should be a legal obligation for the employer to inform the union and the workers, and to consult them, before reducing posts on account of artificial-intelligence applications.
  • Making recruitment algorithms transparent. In preliminary screening done with artificial intelligence, the candidate should have the right to learn the ground and to ask for a human assessment. The KVKK's provisions on automatic decision-making can be used for this.
  • A public obligation for entry-level employment. An obligation, for companies above a certain size, to open paid and secure junior positions in proportion to the number of workers. The labour of young people who are today made to work unpaid, or on low pay, under the name of an internship, should be turned into real employment.
  • Qualified employment in public sectors. Earthquake safety, public health, education, climate adaptation and public transport are the fields that need qualified white-collar labour most. Directing a part of the resources that flow to defence into these fields would be an answer both to unemployment and to social need.
To organise
  • Talk with your graduate friends. The graduate who cannot find a junior position thinks he is alone; yet this is the common condition of a generation. Graduate solidarity networks, and the discussion of this subject in Whom Is the University For? Artificial Intelligence, Cognitive Surrender and the Class of Thinking, turn individual helplessness into a collective demand.
  • Join a union. Even if there is not, in your branch, a union that has passed the threshold, membership is the precondition of organising. The threshold is passed as the members multiply.
  • Discuss the common ownership of the technology. The problem is not artificial intelligence itself, but in whose hands it is. As we asked in For the Restitution of the General Intellect: The Socialist AI Manifesto Is Out: In whose hands, under whose control, to whose benefit? The same technology, under social property, can produce not unemployment but free time.

Dear comrades,

In 1847, answering Proudhon, Marx argued that the machine is not in itself an economic category, and that to understand it one has to look at the social relations inside which it is used. To judge the machine "good" or "bad" is in vain; the real question is who uses it, inside which relation, and to whose benefit.

One hundred and seventy-nine years later, we are inside the same debate. Artificial intelligence is neither saviour nor catastrophe. It is a machine born of the white-collar worker's own accumulation, now stood up against him as capital. If it can have a job done in half the time, there are two paths: either everyone works for half the time, or half of them remain unemployed. Capitalism chooses the second path every time. What will change this is not a technological innovation, but an organised demand.

The poverty of the white collar is learning that expertise does not mean security. But this learning can at the same time be the beginning of a liberation. The diploma, language, the certificate and the career plan promised an individual security, and this promise has been broken. What remains is the only security that has always worked in history: one another.

Dear young comrades, if one thing is to remain from this essay, let it be this:

The poverty of the white collar is not destitution, but the loss of control: losing control over getting a job, over staying in the job, over the shift, over the wage and over the data. Artificial intelligence did not create this loss; it accelerated it. The problem is not that artificial intelligence takes the work; it is that the time it creates returns as profit to capital and as unemployment to labour. The white collar can change this not by individual career strategies, but by organising, in the same rank as the blue collar. The machine is new, the chain is old; and what breaks the chain is not the algorithm, but the organisation.


Sources

News and the interview
Figures from Turkey
Union research and unionisation
Concrete examples
The wage, working time and surveillance
  • TMMOB, Proposal for the Minimum Wage of Engineers, Architects and City Planners for 2026
  • Hümeyra Taşçıoğlu, a study on remote and hybrid work in the computing sector, İmgelem, 20 October 2026
  • Personal Data Protection Board, Principle Decision no. 2026/2035, 16 September 2026
International figures
White-collar organisation in the world
Theory
  • Karl Marx, The Poverty of Philosophy (1847)
  • Karl Marx, Capital, Volume I, the chapter "The Working Day", Part Four (machinery and large-scale industry; relative surplus-value) and Chapter 23 (the general law of capitalist accumulation, the industrial reserve army)
  • Harry Braverman, Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century (1974)
  • Erik Olin Wright, Classes (1985)
Our linked essays

Knowledge Commons - Knowledge Belongs to Everyone. The figures have been compiled from publicly available sources as of 27 October 2026. Michael Page's figure of 32 per cent is the company's own recruitment observation; it is not an official statistic. In the examples of DeFacto Teknoloji and the Community Volunteers Foundation, the parties' claims have been conveyed on the basis of news sources; the proceedings are continuing. CC BY-SA 4.0.

Related Posts

Newsletter