Artificial intelligence was supposed to be the technology that reshaped the global economy. Investors certainly thought so: AI startups pulled in more than $44 billion in the first half of 2025 alone, already outpacing all of 2024. Goldman Sachs now predicts that total investments could hit nearly $200 billion by year’s end.
Yet the optimism may be outpacing reality. A new report from MIT, The GenAI Divide: State of AI in Business 2025, paints a sobering picture. As first reported by Fortune, the study reveals that 95 percent of business attempts to integrate generative AI are failing. Only 5 percent of companies have managed to achieve meaningful revenue acceleration.
The findings fuel growing concerns that the AI gold rush may be inflating a dangerous bubble.
Expectations meet underwhelming performance
For years, Wall Street has banked on AI delivering historic boosts in labor productivity. But the MIT researchers found little evidence that AI is moving the needle in that direction. Previous tests show even the most advanced AI products successfully complete only about 30 percent of assigned office tasks. By April 2025, so-called “AI agents” — hyped as autonomous digital workers — could finish just 24 percent of real-world jobs, according to Fortune.
This gap between hype and performance is already forcing companies to rethink strategy. Gartner’s survey of 163 business executives found that half have abandoned plans to dramatically cut customer service staff by 2027. Kathy Ross, Gartner’s senior director of customer service and support, noted that “the human touch remains irreplaceable in many interactions,” suggesting that companies are learning AI cannot simply replace people overnight.
Workers skeptical, companies backpedaling
Employees appear to share that sentiment. Research from GoTo and Workplace Intelligence found that 62 percent of workers believe AI is “significantly overhyped.” Many IT managers admit their organizations lack formal AI adoption strategies, with security and integration challenges topping the list of barriers.
The mismatch between promises and delivery has already caused public setbacks. Klarna, the fintech company that cut nearly a quarter of its workforce in 2024 in anticipation of an AI-led future, reversed course earlier this year and launched a recruitment drive to rehire staff.
Tech critic Ed Zitron captured the mood bluntly: AI “agents” often sound like intelligent assistants but amount to “trumped-up automations” that still require significant programming effort from businesses.
A new kind of bias: AI choosing itself over humans
Adding another wrinkle, researchers publishing in the Proceedings of the National Academy of Sciences (PNAS) recently discovered a startling trend they call “AI–AI bias.” Large language models like GPT-4 and Meta’s Llama 3.1 consistently favored content created by other AIs over human-written material across product ads, academic abstracts, and even movie reviews.
Study coauthor Jan Kulveit warned that such bias could reshape economic opportunities, with humans at risk of being systematically sidelined. “Being human in an economy populated by AI agents would suck,” he said on X, advising people to run their work through AI tools before submitting it if they suspect another AI will be evaluating it.
This creates a troubling picture: not only are AI systems struggling to deliver promised productivity gains, but they may also be reinforcing their own dominance at the expense of human contributions.
A bubble waiting to burst?
The financial stakes are enormous. Analysts have long projected that AI could add more than $6 trillion to the global economy by 2030. For major tech companies alone, expectations hover around an additional $600 billion in annual revenue — a number so lofty that anything less risks being branded a failure.
Every year AI falls short, the pressure on future productivity gains grows, raising fears that the bubble is unsustainable. As MoneyWeek put it, with so much money riding on AI, only a complete upheaval of existing systems would justify the frenzy of investment.
The AI industry now stands at a crossroads. While breakthroughs remain possible, the MIT study and related research underscore that real-world impact is lagging far behind investor expectations. If productivity does not rise sharply soon, the “AI miracle” may prove to be a mirage.
For businesses and workers alike, the next few years will reveal whether AI is truly a revolutionary force or simply the latest bubble inflated by Silicon Valley optimism.
Yet the optimism may be outpacing reality. A new report from MIT, The GenAI Divide: State of AI in Business 2025, paints a sobering picture. As first reported by Fortune, the study reveals that 95 percent of business attempts to integrate generative AI are failing. Only 5 percent of companies have managed to achieve meaningful revenue acceleration.
The findings fuel growing concerns that the AI gold rush may be inflating a dangerous bubble.
Expectations meet underwhelming performance
For years, Wall Street has banked on AI delivering historic boosts in labor productivity. But the MIT researchers found little evidence that AI is moving the needle in that direction. Previous tests show even the most advanced AI products successfully complete only about 30 percent of assigned office tasks. By April 2025, so-called “AI agents” — hyped as autonomous digital workers — could finish just 24 percent of real-world jobs, according to Fortune.
This gap between hype and performance is already forcing companies to rethink strategy. Gartner’s survey of 163 business executives found that half have abandoned plans to dramatically cut customer service staff by 2027. Kathy Ross, Gartner’s senior director of customer service and support, noted that “the human touch remains irreplaceable in many interactions,” suggesting that companies are learning AI cannot simply replace people overnight.
Workers skeptical, companies backpedaling
Employees appear to share that sentiment. Research from GoTo and Workplace Intelligence found that 62 percent of workers believe AI is “significantly overhyped.” Many IT managers admit their organizations lack formal AI adoption strategies, with security and integration challenges topping the list of barriers.
The mismatch between promises and delivery has already caused public setbacks. Klarna, the fintech company that cut nearly a quarter of its workforce in 2024 in anticipation of an AI-led future, reversed course earlier this year and launched a recruitment drive to rehire staff.
Tech critic Ed Zitron captured the mood bluntly: AI “agents” often sound like intelligent assistants but amount to “trumped-up automations” that still require significant programming effort from businesses.
A new kind of bias: AI choosing itself over humans
Adding another wrinkle, researchers publishing in the Proceedings of the National Academy of Sciences (PNAS) recently discovered a startling trend they call “AI–AI bias.” Large language models like GPT-4 and Meta’s Llama 3.1 consistently favored content created by other AIs over human-written material across product ads, academic abstracts, and even movie reviews.
Study coauthor Jan Kulveit warned that such bias could reshape economic opportunities, with humans at risk of being systematically sidelined. “Being human in an economy populated by AI agents would suck,” he said on X, advising people to run their work through AI tools before submitting it if they suspect another AI will be evaluating it.
This creates a troubling picture: not only are AI systems struggling to deliver promised productivity gains, but they may also be reinforcing their own dominance at the expense of human contributions.
A bubble waiting to burst?
The financial stakes are enormous. Analysts have long projected that AI could add more than $6 trillion to the global economy by 2030. For major tech companies alone, expectations hover around an additional $600 billion in annual revenue — a number so lofty that anything less risks being branded a failure.
Every year AI falls short, the pressure on future productivity gains grows, raising fears that the bubble is unsustainable. As MoneyWeek put it, with so much money riding on AI, only a complete upheaval of existing systems would justify the frenzy of investment.
The AI industry now stands at a crossroads. While breakthroughs remain possible, the MIT study and related research underscore that real-world impact is lagging far behind investor expectations. If productivity does not rise sharply soon, the “AI miracle” may prove to be a mirage.
For businesses and workers alike, the next few years will reveal whether AI is truly a revolutionary force or simply the latest bubble inflated by Silicon Valley optimism.
You may also like
Eberechi Eze: Crystal Palace make late U-turn after agreeing Arsenal transfer
Last One Laughing new series update as Bob Mortimer returns to defend title
Millie Bobby Brown announces she's welcomed baby girl by adoption in sweet update
Canary Wharf fire LIVE: Thick smoke pouring from London tower block as flames seen inside
Bank boosts interest on Cash ISA to 4.21% and earns 'excellent' rating
