In the early 1470s, a Benedictine monk named Filippo de Strata sat down to write a furious letter to the Doge of Venice. Printing had just arrived in the city, and Filippo was not taking it well. His polemic was not simply a plea for scribal jobs. It was a lament for a civilization. The scribe, he argued, was not a copyist. The scribe was the keeper of knowledge itself. To replace him with a machine was not progress. It was an assault on what it meant to matter.

Filippo was defending a monopoly. For four thousand years, from Mesopotamian cuneiform clerks to medieval guild scribes, literacy had been the defining cognitive advantage of a small professional class. To read was to hold power. To write was to constitute reality: land records, contracts, tax rolls, sacred texts. The scribe was simultaneously the HR department, the finance department, the legal department, and the public record of his civilization.

Then came the machine that made their skill ordinary.

This is not a story about technology. It is a story about what happens to human identity when a capacity that previously defined significance becomes, almost overnight, cheap and common. The scribe’s crisis was not an employment crisis in the first instance. It was an identity crisis. And it is the same crisis, wearing different clothes, that now confronts hundreds of millions of knowledge workers in 2025.

A different kind of machine

Every prior wave of cognitive automation targeted narrow, specific tasks. The calculator replaced mental arithmetic. OCR replaced data entry clerks. Search engines replaced the reference librarian’s function of knowing where things were. Each tool attacked one skill.

Large language models do something categorically different. They do not automate a task. They automate a domain. The same system that drafts a legal brief can write marketing copy, generate code, summarize a clinical trial, and handle a customer conversation. For the first time in the history of tools, what is being externalized is not a single human capability but something closer to the general capacity for language-mediated reasoning, which is the foundation of almost every white-collar profession.

This is the rupture Filippo could not have imagined: not a machine that copies faster, but one that composes, reasons, interprets, and advises. And it does so at a cost that approaches zero per query.

The economic signals are directionally hard to dismiss. Hiring for white-collar positions in the US reached decade lows in 2024. Approximately forty percent of white-collar job seekers in that year failed to secure a single interview. Cloud computing and systems design industries stopped growing at the precise moment ChatGPT launched. The revolution does not announce itself with shuttered factories. It arrives through hiring freezes and shrinking entry-level cohorts.

Who pays the price

The question of who bears the cost of disruption is never separate from who holds power when the disruption begins.

In the industrial revolution, women entered the factory at lower wages, and domestic labour remained entirely unpaid and entirely female. The digital revolution created an attention economy that monetized data without compensating those who generated it, a dynamic that fell hardest on those with least leverage over their own information. The AI disruption is repeating these patterns with new technical surfaces.

Women are disproportionately concentrated in the clerical and routine cognitive roles that generative AI can most readily automate. Research across eighteen studies covering roughly 143,000 people found a consistent twenty-five percent gap in AI adoption between men and women. Women who reported lower adoption often cited fears of professional penalties for perceived inauthenticity, even when their competence with the tools was no different. This is the same double bind from every prior disruption: higher reputational penalties for using shortcuts, combined with higher automation exposure because they dominate the roles the shortcuts displace.

A Harvard Business Review study found that female engineers using AI for code generation were rated nearly nine percent less competent than male engineers, even when evaluators were looking at identical outputs. The machine produces the work. The human is still judged by gender. Fewer than thirty percent of the AI workforce is female, dropping to fifteen percent at leadership levels. The people building the systems that will reorganize whose work is valued are overwhelmingly men. This is not incidental. It is structural.

India’s specific stakes

For India, the stakes are particularly sharp. The BPO and IT export sector, a $200 billion industry employing close to four million people, was built on a clear promise: cognitive labor translates into social mobility. The educated graduate in Chennai or Hyderabad was not defined by her hands. She was defined by her ability to navigate English-language systems, reason through problems, produce reliable analytical outputs. Her education was her entry into the middle class.

Generative AI is now automating the same tasks that constituted that mobility. In 2024, PhonePe laid off sixty percent of its customer support staff, replacing human agents with AI-driven systems. Market analyses estimate that over seventy percent of Indian BPO firms now leverage AI or natural language processing in their operations. An EY India report found that call center management faces an eighty percent productivity enhancement from AI, which, in this context, is a polite term for workforce reduction.

A 2024 IIM Ahmedabad study found that sixty-eight percent of white-collar employees expect AI to partially or fully automate their jobs within five years. This in a country where the ILO already documented that educated youth unemployment doubled between 2000 and 2022, before generative AI arrived at scale.

There is a secondary dynamic that rarely surfaces in mainstream analysis. The workers who annotate data, label images, and correct AI outputs, the hidden cognitive infrastructure of systems now displacing their colleagues, are disproportionately located in the Global South. They earn dollars per hour. The AI systems they train are worth billions. AI appears intelligent partly because human beings, mostly poor ones, made it so.

The crisis beneath the crisis

The deepest risk of the AI transition is not that machines will take jobs. It is that they will hollow out the particular form of significance that cognitive work has provided since the industrial era.

For a hundred and fifty years, educated labor has told itself a coherent story: my value lies in what I can reason, draft, analyze, judge. That story organized degrees, career trajectories, and the social respect that accrues to someone who “works with their mind.” It organized the middle class.

That story is now being commoditized. When a law partner’s first draft and a language model’s first draft are indistinguishable to most readers, what remains of the distinction that mattered? This is not abstract philosophy. It is being lived in real time by early-career professionals discovering that entry-level roles, once cognitive apprenticeships, are contracting, and by mid-career professionals sensing their expertise is devalued before they have repaid the debt incurred acquiring it.

Historically, the disruptions that produced the most lasting damage were those where transition costs fell entirely on the least powerful, without social contract, without redistribution. The conversations society must now have require naming what is actually at stake. It is not productivity. It is dignity. And the costs will not be evenly distributed. They will follow the operating systems of gender, caste, class, and geography that have structured every prior disruption.

Filippo de Strata was wrong about the printing press. But he was asking the right question. Not, can we stop this? But, what do we owe each other in the passage through it?

END OF ARTICLE