Economists noticed something perplexing almost forty years ago, when personal computers were still a novelty humming away in beige boxes throughout corporate America. Businesses were investing a staggering amount of money in these machines, but worker productivity, which everyone thought would skyrocket, hardly changed. It was dubbed the “productivity paradox.”
Both academics and business executives were genuinely perplexed by it, and there was a years-long collective scratching of heads. The same paradox has now reappeared, albeit in a different guise. It has a generative AI face this time.
| Category | Details |
|---|---|
| Total Global AI Investment (2013–2024) | $1.6 trillion — per Stanford University’s 2025 AI Index Report |
| Projected Worldwide AI Spending (2026) | $2.5 trillion (44% increase over 2025), according to Gartner |
| Business AI Adoption Rate | ~80% of companies report using generative AI |
| Companies Reporting No Bottom-Line Impact | Nearly 80%, per McKinsey & Co. research |
| AI Pilot Projects Abandoned by End of 2024 | 42% of projects scrapped — up from 17% the year before |
| Expected Business Investment in Gen AI (This Year) | $61.9 billion — a 94% increase, per IDC |
| Top AI-Spending Country | United States — approximately 62% of total private AI funding since 2013 |
| US Corporate AI Spend (2013–2024) | $471 billion |
| Second and Third Largest AI Spenders | China ($119bn), United Kingdom ($28bn) |
| Major AI Infrastructure Spenders | Microsoft, Meta, Alphabet, Amazon, Oracle — projected $500B–$700B+ combined in 2026 |
| JPMorgan AI Users | 200,000 employees with access to AI assistant; ~half use it regularly |
| USAA Daily Call Volume | 200,000 calls per day handled with AI-assisted human agents |
| Gartner’s Current AI Cycle Stage | Sliding toward “trough of disillusionment” — low point expected in 2025–2026 |
| Key Failure Reasons | Technical hurdles AND human factors — employee resistance, lack of skills |
Nowadays, you’ll hear the term “AI” in every other meeting at practically every large corporate headquarters. It’s being pitched by consultants. Boards are calling for it. Investors are leaning forward in their seats as CEOs confidently promise it during earnings calls. Nearly eight out of ten businesses have implemented generative AI in some capacity, according to McKinsey & Co.
However, nearly the same percentage say their bottom line has not significantly improved. Eight out of ten will embrace it. Eight out of ten are waiting for it to function. That symmetry has a darkly humorous quality.
The figures being presented are astounding in a way that is truly hard to understand. The Apollo Program, the US Interstate Highway System, and the International Space Station put together pale in comparison to the $1.6 trillion global corporate AI investment between 2013 and 2024. Businesses are predicted to invest $61.9 billion in generative AI this year alone—nearly twice as much as they did the previous year.
According to Gartner, global AI spending is expected to reach $2.5 trillion by 2026. Nevertheless, by year’s end, 42% of AI pilot projects had been shelved, a significant increase from 17% the previous year. The money is coming in. The outcomes aren’t.
It’s possible that expectations simply outpaced reality too quickly. The ChatGPT moment in late 2022 was like a thunderclap: all of a sudden, a machine could produce code, write legal memos, and provide polished prose responses to customer inquiries.
It was easy to extrapolate: if it can do this now, what about next year? Businesses didn’t wait to envision. They began to spend. However, the difference between what a conference room demo suggests and what a deployment across 50,000 employees actually delivers proved to be significant and fraught with difficulties.
S&P Global senior analyst Alexander Johnston identified “human factors” as one of the most enduring issues. Workers oppose tools they don’t trust. When they feel like software is handling them, customers push back. Additionally, the learning curve for incorporating AI into actual workflows in some industries is higher than that of any vendor that was ever mentioned in the pitch deck. Cultural barriers may be more significant than technical ones.
Because it demonstrates how cautiously even a willing company must move, rather than because it is a failure, USAA provides an instructive case study. The 16,000 customer service representatives of the military-focused insurance and banking company now have access to an AI assistant that helps them find precise answers when making calls. It decided on a tool that enhances rather than replaces humans after conducting pilot projects, some of which it completely shut down. 200,000 calls are still handled daily by its call centers.
In short, people want a human voice on the other end of the phone, according to Ramnik Bajaj, chief data analytics officer at USAA. That is not an AI failure. It serves as a reminder that technology is not isolated.
A different, and possibly the most accurate, picture of what widespread AI adoption actually looks like in practice is provided by JPMorgan Chase. The bank completely blocked ChatGPT from its systems two years ago. Currently, an internal AI assistant is used by about 200,000 workers. Up to four hours a week are saved on routine tasks by half of them who use it on a regular basis. The bank’s global chief information officer, Lori Beer, is in charge of 60,000 employees in the technology department.
She freely acknowledges that the bank has terminated hundreds of AI projects. That doesn’t seem to bother her. “We’re not afraid to shut things down,” she replied. “I think it’s a smart thing.” There’s a feeling that JPMorgan is one of the few non-tech companies that appears to be making real progress because of its willingness to fail quietly and move on.
Almost all of the winners thus far have been on the supply side. Instead of waiting for results, Microsoft, Amazon, Google, and Nvidia are the outcomes. Nvidia has become one of the most valuable companies in human history thanks to its dominance in AI chips. When Meta showed that its AI investments were significantly increasing its advertising revenue, its stock shot up.
Conversely, Microsoft’s stock plummeted when Azure growth fell short of projections, creating uneasy doubts about whether all that infrastructure investment will result in profits that investors can truly hang onto. The returns are inconsistent even within Big Tech, and Wall Street’s tolerance is clearly waning.
The trough of disillusionment is Gartner’s term for where technology is headed in the near future. It’s a sobering reset rather than a dramatic collapse, similar to what happened with personal computers and the internet in the early 2000s. Early exuberance, costly confusion, gradual mastery, and ultimately transformation are all part of the same pattern.
According to MIT research scientist Andrew McAfee, who focuses on digital economies, “Innovation is a process of failing fairly regularly.” For a CFO attempting to explain AI spending to a skeptical board, this is not a reassuring message. However, it could be an honest one.
It’s difficult to ignore the fact that the businesses promoting AI are frequently the ones who are most outspoken about its potential. The technology is not a scam despite this. However, it does mean that the businesses making the purchases bear nearly all of the cost of the gap between the pitch and the payout. They are investing billions, learning from their mistakes, educating their employees, and biding their time.
The paradox is not that AI is ineffective. It’s that, contrary to what the headlines ever implied, making it work is much more difficult, slower, and human.
