Office jobs vs manual labor: who survives the AI revolution?

For years, automation was expected to eliminate factory jobs first. But with AI replacing junior coding roles faster than manual labor, the future of work is shifting.

Nearly a decade ago, I spent six months working in a factory in Iceland. It was a tedious job involving machine monitoring and manual labor, which, as I overly optimistically thought, would be replaced by automation within 10-15 years.

After I returned to Lithuania, my home country, I took a short coding course, thinking of a career that would guarantee a job for a lifetime in a profession that’s always in demand.

Years later, millions of humans are still working in factories, while AI is replacing junior coders due to breakthroughs in the development of large language models (LLMs).

Meta’s CEO, Mark Zuckerberg, recently announced that the company will start replacing its mid-level software engineers as soon as this year.

The CEO of Replit, a company providing tools for building websites, said that it doesn’t need professional coders thanks to its AI-enabled product, which can write software applications from a prompt. Many companies already offer similar solutions, allowing one to build websites and apps without technical knowledge.

In contrast, so-called blue-collar positions, including manufacturing and construction, are automated at a much slower pace.

Will the breakthrough of LLMs result in blue-collar jobs having a much longer lifespan, or will robotics catch up? Do programmers and other office workers face extinction?

Less boring work for software engineers

Devansh Agarwal, a senior machine learning engineer at Amazon Web Services who has written several research papers on AI, believes that AI won't ever replace software engineers. Instead, it will greatly augment them and reduce the amount of "boring" work.

Software engineers will be in demand, he says, as their work is complex and includes many tasks, such as maintaining and reviewing code, designing documents, coming up with algorithms, and debugging production issues.

According to Agarwal, even the largest LLMs with more than 600 billion parameters are not yet able to do one task properly. He doubts that models with even 10x parameters could replace software engineers, as there is not enough data to learn from.

“There is more to this problem. To get a decent output from an LLM, we need to try multiple prompts, which adds up to the cost. It is very hard to provide all the product requirement information to an LLM and expect it to generate a solution as the product requirements are changing all the time,” Agarwal says.

He expects LLMs' impact on coding to be similar to that of software engineers. However, current market trends already show that fewer jobs require only coding, signaling that programmers will have to expand their abilities with other skills.

Coders trust AI too much

Mantas Lukauskas, the AI Tech lead at web hosting company Hostinger, says that AI has the biggest impact on junior positions in technical fields such as coding and data analysis.

“Previously, you needed to understand what to calculate, but now you can simply upload a data table to ChatGPT, and it will provide all the information – you no longer need an entry-level data analyst,” notes Lukauskas, who is also responsible for AI development in AI orchestration startup nexos.ai and is a researcher at the Kaunas University of Technology.

A simple coding job such as creating a website can be done by companies such as Vercel, allowing one to write code without technical skills.

The rise of such coding tools and AI assistants like Github Copilot will require future programmers to learn new skills.

“There is creativity, then general computer literacy, followed by AI-related skills. Critical thinking is essential as well – many people trust LLMs too much. They generate code and simply copy it without even checking if it's optimal,” Lukauskas notes.

Beyond coding, AI is impacting many other positions. According to Lukauskas, Hostinger chatbots already handle around 50% of the workload, while AI-based tools enhance productivity by eliminating repetitive tasks and allowing humans to focus on more strategic work.

Coding is not obsolete

A recent Anthropic Economic Index, which uses anonymized interactions with Claude.ai chatbot, offers insights into which industries could be reshaped by AI at the fastest pace.

Beyond computer and mathematical positions, which had the highest associated AI usage rate (37%), arts, design, entertainment, sports, and media occupations were second most common (10.3%). The authors note that this likely reflects the prevalence of marketing, writing, and other kinds of content.

Natalie Lambert, founder and managing Partner at GenEdge Consulting, says that AI is already handling content creation, basic design, data analysis, and competitive intelligence in marketing.

“I foresee a near future where marketing teams achieve the same output with at least 50% fewer people,” she forecasts.

She illustrates AI’s impact on coding by drawing on her personal experience. Creating a website with AI tools now costs as much as a monthly subscription, whereas businesses spent thousands doing the same thing just a year earlier.

However, Lambert thinks that the growing capabilities of AI don’t mean that coding is obsolete. New tools will enable developers to focus on more complex and high-value work, such as security and custom functionality.

“The challenge is that while demand for advanced developers is growing, junior and mid-level developers may struggle to find opportunities. That raises an important question: how do you become an advanced developer without starting at the beginning?” she says.

Coding may become interdisciplinary

Severine Zaslavski, Vice President of Global Product Development at ManpowerGroup, thinks that coding still remains a strong career choice for newcomers.

“Newcomers should look into emerging areas like low-code/no-code platforms, cloud-native development, edge computing, quantum computing, and explore niche skills such as ethical AI practices,” she says.

“Foundational skills such as analytical thinking, mathematical proficiency, adaptability, and curiosity are critical for thriving in the rapidly evolving job landscape. ”

Manpower expects that within 20 years, coding will become more interdisciplinary, combining expertise from specific sectors such as biology, neuroscience, and other fields to tackle complex challenges.

As for other white-collar professions, automation may reduce demand in office support and customer service roles, though it is more likely to enhance STEM, creative, and business professions by boosting productivity rather than outright replacing them, Zaslavski says.

Anthropic Index also notes that AI’s use leans more toward augmentation (57%) – where AI collaborates with and enhances human capabilities – compared to automation (43%).

Can robotics catch up?

While AI significantly impacts some office jobs, its current impact on so-called blue-collar positions is far smaller.

According to the Anthropic Index, occupations requiring physical labor use AI the least. These include transportation and material moving, healthcare support, farming, fishing, and forestry.

Lukauskas notes that companies rushed to implement robotic solutions years ago, and many feared that the automation revolution would replace their jobs.

“Everything has changed since LLMs emerged. Even though they might influence decision-making in robotization, they do not affect the physical world.”

Agarwal, who also advises at a robotics startup, thinks that manual labor jobs are at a higher risk than white-collar jobs, as it is becoming easier to program robots using GenAI.

Zaslavski says that in the near term, jobs requiring hands-on skills, physical presence, or human interaction – such as construction workers, delivery drivers, personal care aides, and healthcare providers – are less susceptible to automation.

“These roles rely on adaptability and dexterity in dynamic environments, making them harder to replace entirely with AI or robotics. Similarly, roles that integrate advanced technology, such as renewable energy engineers or autonomous vehicle specialists, are future-proof as they align with evolving trends in green energy and infrastructure development.”