I’ve been wondering about this for a long time, but is the claim that AI will take away jobs really valid? When I read an article written by a partner at a16z, they argued that the premise itself is flawed.

The basic misunderstanding seems to be the idea that “the total amount of work that should be done worldwide is fixed.” But when you look at history, that has never happened. In fact, it’s the opposite.

Take agriculture as an example: in the early 20th century, one-third of the labor force in アメリカ was engaged in agriculture. As tractors and mechanization advanced, that share fell to 2% by 2017. Even though employment was expected to completely disappear, agricultural production actually tripled, and those workers flowed into factories, stores, offices, hospitals, and eventually even the IT industry.

The same is true for the electrification era. In the early 1900s, only 5% of factories in アメリカ were using electricity. By 1930, 80% were running on power. Even though productivity doubled, unemployment did not increase. Instead, new industries kept being created one after another.

When Excel came out, people said, “Bookkeeping jobs are over.” Sure, the number of bookkeepers decreased by 1 million. But 1.5 million more financial analysts appeared. It wasn’t that employment disappeared—jobs just changed in quality.

Looking at current data on AI, it’s actually not as bad as doom-mongers claim. After multiple research institutions examined it, they reported that about 95% of companies that adopted AI saw no impact on their headcount. Even among the companies where there was an impact, increases and decreases were almost the same.

In Goldman Sachs’ analysis, they say the effect of AI is far greater for “enhancing” jobs than for “substituting” them. In earnings calls, the phrase “functional enhancement” was used about 8 times as often as “substitution.”

The surge in demand for software engineers is also symbolic. Even though AI is making coding more efficient, the number of engineering job openings is increasing instead. GitHub push counts are exploding, and new applications are exploding too. Demand for product managers has also reached its highest level since 2022.

The example of travel agencies is also interesting. In 2000, travel agency employees were earning 87% of the average salary across all industries. Even though technology reduced that by half, by 2025 that share is expected to rise to 99%. In other words, jobs may have decreased, but the people who remained are earning higher wages.

What matters here is that most of the jobs created after 1940 were job categories that didn’t even exist in 1940—cloud engineers, data scientists, social media managers, robotics-related roles, and more. There are countless jobs that people at the time couldn’t have even imagined.

AI might indeed replace some routine administrative work and management tasks. But at the same time, advanced work like analysis, technology, and management is strongly more likely to be enhanced and complemented. On a macro level, there’s also data showing no statistically significant correlation yet between AI and unemployment rates.

For example, robotics-related datasets jumped from 10th place to 1st in just 2 years. Before AI fully penetrates society, even in this field alone there are enormous unmet needs for work that still hasn’t been filled.

In the end, human desires have no limits. When food becomes cheaper, people spend money on housing, healthcare, education, travel, and entertainment. The same applies to jobs. As productivity rises, new demand emerges, and new job titles appear. This is proven by the economic history of the past 200 years.

The argument that AI will end employment only holds if human needs disappear the moment AI becomes cheaper—which is impossible. Instead, cheaper AI, larger markets, new companies, new industries, and more complex human jobs are what should be coming next.