AI and the Future of Work in IT: What J. Keynes Got Right About Automation and Productivity

Almost a century ago, John Maynard Keynes predicted that technological progress would eventually solve humanity’s problem of scarcity. Instead of working to survive, future generations, he believed, would work to fulfill themselves.

Today, as artificial intelligence begins to automate not just physical labor but cognitive work, including software development, Keynes’s question feels strikingly relevant again. Is AI eliminating jobs, or is it transforming the very structure of work in IT? This article explores how the rise of generative AI echoes past technological revolutions and what it really means for productivity, skills, and the future of the profession.

Keynes’s Prediction: From Scarcity to Abundance

In 1930, John Maynard Keynes, in his essay Economic Possibilities for Our Grandchildren, suggested that humanity’s central economic problem — the struggle for subsistence — was not permanent. Through capital accumulation, scientific progress, and machine-based production, productivity would increase so dramatically that within a century society would face not scarcity, but abundance.

Keynes argued that once productivity reached a sufficient level, work would cease to be primarily a means of survival and would instead become a domain of self-realization and voluntary participation.

The Machine Revolution and Technological Unemployment

Keynes viewed the transition from manual labor to machine-based production as a historical turning point. For the first time, machines systematically replaced human muscle. This led to what he called “technological unemployment” — a temporary imbalance between the speed of automation and society’s ability to create new roles.

However, in the long term, he expected working hours to shrink and the focus of human activity to shift from pure production toward cultural and intellectual development.

The Shift Toward Automating Thought

Today, we are witnessing a similar but deeper transition: from the automation of physical labor to the automation of cognitive labor. Artificial intelligence and “vibe coding” tools are beginning to replace not only manual tasks but also substantial portions of intellectual work, including programming, analysis, and design.

If the industrial era mechanized the body, the AI era mechanizes thinking. This raises the same fundamental question Keynes posed nearly a century ago: if productivity grows faster than the need for human participation, how will employment structures, skill value, and the very nature of work change?

The Paradox of Generative AI in Software Development

The current AI landscape in IT is deeply contradictory. Generative models can produce vast amounts of code, yet that code is often redundant, non-idiomatic, or poorly integrated into project architecture.

Some open-source communities are already facing an influx of pull requests generated fully or partially by large language models (LLMs), requiring more time to review than manual contributions would have required. Many such submissions are rejected, and some projects have even restricted AI-driven “bug hunting.”

In this sense, AI introduces a new form of “technological noise”: quantitative productivity increases, but qualitative output does not always follow.

The Hidden Costs of Replacing Developers with LLMs

In some companies, attempts to replace development teams with LLM-based tools have resulted in hidden costs: growing technical debt, degraded architecture, and more complex maintenance.

Code can be generated quickly. But system understanding, architectural responsibility, and long-term design thinking do not scale at the same linear rate. This fuels anxiety among professionals who see AI not as a tool, but as a destabilizing force in the labor market.

A Recurring Pattern in Technological Revolutions

Viewed historically, what we are experiencing is not a unique crisis but a typical stage of a technological revolution.

Every major technological transformation lowers barriers to entry and increases “surface-level” productivity while simultaneously generating an oversupply of low-quality output. This happened during industrialization, mass production, the rise of personal computers, and the spread of the internet.

The transition period is always marked by overheating expectations, illusions of rapid human replacement, and eventual correction.

AI as a Redistribution of Complexity and a New Infrastructure Layer

Following Keynes’s logic, AI in IT is less about eliminating labor and more about redistributing complexity. Machines begin to automate the routine cognitive layer, while value shifts upward, toward architecture, systems thinking, integration, verification, and risk management.

It is not the disappearance of the profession, but its transformation.

As with previous technological shifts, AI initially triggers anxiety, redistributes roles, and intensifies competition. But in the long run, it expands society’s productive capacity.

In this sense, artificial intelligence can be viewed not as a replacement for specialists, but as a new infrastructure layer, comparable to the steam engine, electricity, or the internet.

The Core Question of Productivity

The fundamental question remains the same as in 1930: will rising productivity reduce routine work and lower burnout, or will it simply intensify expectations and competition among professionals?

The issue is not the technology itself, but whether it enhances human productivity or is used primarily for short-term cost optimization. AI can serve as a cost-cutting mechanism, or it can become an accelerator for research, prototyping, and interdisciplinary integration.

Automation Raises Complexity, It Does Not Eliminate Work

Historically, automation has not destroyed work. It has raised the level of complexity. Machines displaced physical routine labor but created engineering, managerial, and scientific professions.

Similarly, AI automates part of cognitive routine work, freeing space for more complex tasks: systems design, modeling, scientific hypothesis generation, and large-scale architectural solutions.

In this sense, AI may function not as a tool of replacement but as a driver of a new scientific and technological cycle, reducing the time between idea and implementation.

The real question is not whether specialists will disappear, but which competencies will become central in a rapidly evolving professional environment.