What Happened to the AI Revolution?

The Hype Is Over

Six major tech companies — Nvidia, Alphabet, Amazon, Apple, Meta and Microsoft — are heavily investing in artificial intelligence, according to The Economist. This year, they are budgeting about $400 billion in capital expenditures, mainly for AI-related equipment, as well as research and development.

In San Francisco, the world’s "tech capital," it's still believed that AI will transform the global economy. However, for AI to unlock its potential, companies worldwide need to buy the technology, adapt it to their needs, and become more productive as a result. Investors have added more than $2 trillion to the market value of these five major tech companies over the past year, essentially forecasting an additional $300-400 billion in annual revenue. However, as of now, these tech giants are far from achieving such results. Even optimistic analysts estimate that Microsoft will earn only about $10 billion this year from sales related to generative AI. Outside the West Coast of America, there are few signs that AI is having a significant impact on anything, the article notes.

One issue is the speed of adoption. Companies publish striking estimates of how many people use generative AI. Nearly two-thirds of respondents to a recent McKinsey survey said their company "regularly uses" this technology, almost twice as many as a year earlier. A report from Microsoft and LinkedIn states that 75% of "knowledge workers" (people who sit at a computer all day) worldwide use it. According to them, people are already in the AI world.

In a sense, they are right. Almost everyone uses AI when searching for something on Google or choosing a song on Spotify. But implementing AI in business processes remains a specialized task. Official statistical agencies ask firms of all kinds and in a wider range of industries about AI. The U.S. Census Bureau provides the best estimates, finding that only 5% of businesses used AI in the past two weeks. Even in San Francisco, many tech specialists admit when pressed that they don't pay $20 a month for the best version of ChatGPT. Similar stories exist elsewhere. According to official Canadian data, 6% of firms used AI for producing goods and providing services in the past 12 months. British surveys show higher usage — 20% of all enterprises in March — although questions are asked differently. And even in the UK, usage is growing slowly; the same share used AI in September last year.

Concerns about data security, biased algorithms, and AI errors ("hallucinations") slow deployment. McDonald's, a fast-food chain, recently shut down a trial where AI took customer orders after the system began making mistakes, such as adding $222 worth of chicken nuggets to one customer's bill. Additionally, some companies hesitate due to an abundance of pilot projects — too many small AI projects make it difficult to decide where to invest. Other firms refrain from large projects because AI is developing so quickly that it's easy to spend on technologies that will soon become obsolete.

Companies that go beyond experiments use generative AI for a narrow range of tasks. Optimizing customer service is perhaps the most common. ADP, a payroll company, boasts a "new feature that allows small business clients... to use AI to answer questions and better understand how to initiate HR actions." Others use the technology for marketing. Verizon, a telecommunications firm, says it uses AI to create better "personalized plan recommendations" for its customers; Starbucks uses it to create "more personalized offers for customers."

"If you think these efforts seem a bit unimpressive, you're not alone. Goldman Sachs built a stock index tracking companies that the bank believes have 'the most significant potential base profit change from AI implementation through productivity gains.' The index includes companies like Walmart, a major grocery chain, and H&R Block, a tax preparation company. Since the end of 2022, the stock prices of these companies have failed to outperform the broader stock market. In other words, investors do not see prospects for additional profits. Technology might even distract managers from more pressing matters," the article states.

What about stories of companies using AI to transform their operations? Klarna is one often-cited example. The online financial services company recently said its AI assistant performs the work of 700 customer service agents. Its CEO says this technology reduces the company's employment by one-fifth each year. However, this is at best an incomplete picture. Klarna hopes to go public soon, and talk of its AI use attracts media attention. According to consulting firm CB Insights, Klarna's staff began shrinking long before AI appeared. The company is worth perhaps half of its 2021 value. If it's cutting employees now, excessive hiring during the COVID-19 pandemic is as much to blame as AI's merits.

No signs of layoffs are visible in macroeconomic data. Kristalina Georgieva, head of the IMF, recently warned that AI would hit the labor market like a "tsunami." However, unemployment in wealthy countries is currently below 5%, close to a historic low. The share of wealthy-world workers at work is near a historic high. Wage growth also remains high, which is hard to reconcile with an environment where workers' bargaining power is supposedly weakening.

AI's effect also doesn't show up if you dig deeper into the numbers. Workers aren't moving between companies faster than usual, which would likely happen if many jobs were disappearing. Using U.S. employment data by occupation, you can look at various roles ranging from back-office support to copywriters. Such roles are considered vulnerable to AI, which is improving at tasks requiring logical thinking and creativity. Despite this, the share of employment in "white-collar" professions is a percentage point higher than before the pandemic.

Some economists believe AI will transform the global economy without displacing people from jobs. Collaboration with a virtual assistant can increase productivity. In a new paper, Anders Humlum from the University of Chicago and Emilie Westergaard from the University of Copenhagen survey 100,000 Danish workers. The average respondent estimates that ChatGPT could halve the time spent on about a third of work tasks, which theoretically is a significant efficiency gain.

However, macroeconomic data also shows no substantial evidence of productivity growth. Recent estimates based on official data show that real output per worker in the average wealthy country is not growing at all. In America, the global AI hub, output per hour remains below its pre-2020 trend. Even in global survey data from purchasing managers, which is produced with a shorter lag, there are no signs of productivity growth.

For such a leap, firms need to invest in AI. Except for major tech companies, which mostly spend on developing AI products for others rather than boosting their productivity, most companies are not doing this. Capital spending among other S&P 500 companies, the index of America's largest companies, is likely to decline in real terms this year. Overall, business investment in equipment and software for information processing is growing at an annual rate of 5%, well below the long-term average. In the entire wealthy world, investment is growing slower than in the 2010s.

Over time, companies may realize AI's true potential. Most technological waves, from tractors and electricity to personal computers, take some time to spread through the economy. Indeed, assuming that major tech companies' AI revenues grow at an average annual rate of 20%, investors expect nearly all AI revenues from major tech companies to come after 2032.