Something is breaking in the AI narrative.

Not quietly. Not subtly. The data is now screaming what a few of us have been saying for over a year: the automation-first model of AI — what I call AI 1.0 — is failing at enterprise scale. And the market is starting to notice.

Let me walk you through what's happening. I'm calling it "The Shift."

The Collapse of AI 1.0 by the Numbers

Yesterday, Fortune reported on a new NBER study of 6,000 CEOs and CFOs across the US, UK, Germany, and Australia. The finding: 90% of firms reported AI had zero impact on employment or productivity over the past three years. Economists are now openly comparing this to Solow's Productivity Paradox from 1987: "You can see the computer age everywhere but in the productivity statistics."¹

Replace "computer" with "AI" and you have 2026.

But it gets worse:

The NBER study found that despite 90% of firms reporting no productivity impact, executives still project AI will boost productivity by 1.4%. Hope is not a strategy.³

Microsoft Copilot — the flagship enterprise AI product, backed by $60M+ in TV advertising — has penetrated just 3% of its 450 million paid seats. Gartner found only 6% of enterprises have moved generative AI beyond pilot stage.⁴

S&P Global reports 42% of companies scrapped most of their AI initiatives in 2025, up from 17% the year before. That's not a dip. That's a trend reversal.⁵

ManpowerGroup's 2026 Global Talent Barometer: workers' AI usage increased 13%, but confidence in the technology's utility dropped 18%. People are using it more and trusting it less.⁶

Deloitte surveyed 3,235 leaders reimagining their business with AI. Just 20% report actual revenue growth. 74% still "aspire" to it.⁷

Only 1% of organizations consider their AI practice mature, despite 78%+ reporting active use.⁸

Read that again: 78% using it. 1% mature. That's not an adoption problem. That's a paradigm problem.

The Financial Markets Are Waking Up

The Bank of England has warned of growing risks of a global market correction driven by AI overvaluation. The IMF drew explicit comparisons to the dot-com bubble.⁹ The World Economic Forum published a detailed timeline of how an AI bubble burst would unfold.¹⁰

One top economist noted that three of his four "horsemen" of a bubble — overvaluation, bubble beliefs, and investor inflows — are already present. The only missing piece? A wave of IPOs. OpenAI is reportedly planning to go public in Q4 2026.¹¹

When the last horseman arrives, pay attention.

Meanwhile, enterprises are buying fewer AI licenses, with a growing sentiment that vendors are "forcing" them to adopt costly AI features that should be included in base platform fees.¹²

The Shift: From AI 1.0 to AI 2.0

Here's what almost no one is talking about: the reason AI 1.0 is failing isn't because AI doesn't work. It's because the fundamental premise is wrong.

AI 1.0 asked: "How do we replace humans with machines?"

AI 2.0 asks: "How do we make humans more capable?"

This isn't just my thesis anymore. The data is converging from every direction:

TechCrunch's 2026 outlook declared the industry is "sobering up," predicting that "AI has not worked as autonomously as we thought" and that the conversation will shift to augmenting human workflows. One expert said outright: "2026 will be the year of the humans."¹³

Stanford's Erik Brynjolfsson argues that using AI to automate human intelligence is "an incredibly powerful and evocative vision, but it's a very limiting one." The much larger opportunity lies in augmentation.¹⁴

A Stanford SALT Lab study of 1,500 workers across 104 occupations found that workers overwhelmingly prefer to retain oversight, judgment, and control when working with AI. They want augmentation, not replacement.¹⁵

The California Management Review published a framework arguing that "sustainable competitive advantage will stem not from wholesale automation, but from orchestrating thoughtful transitions that recognize both technological potential and human value."¹⁶

Multiple 2026 industry panels now declare "the narrative has changed" — from automation to augmentation.¹⁷

A mid-market manufacturing COO captured the disconnect perfectly in the MIT study: "The hype on LinkedIn says everything has changed, but in our operations, nothing fundamental has shifted."¹⁸

The Implications

Here's why this matters:

1. The $1.1 trillion bet is misallocated. Mega-cap AI spending is projected at $1.1 trillion between 2026–2029.⁹ Most of it is being spent on infrastructure designed for the automation paradigm. When The Shift becomes undeniable, the companies that repositioned early will capture disproportionate value.

2. The 95% failure rate is a feature, not a bug. These pilots aren't failing because of bad execution. They're failing because they're trying to automate processes that require human judgment, creativity, and contextual understanding. You can't optimize your way out of a paradigm problem.

3. The augmentation economy is nascent but inevitable. Companies using AI-human collaboration are reporting 40–60% productivity gains (BCG & Wharton) and 2.3x higher revenue growth (Deloitte) compared to automation-only approaches. The ROI isn't close.¹⁹

4. The workforce is already voting. The shadow AI economy — where 90% of workers use unsanctioned tools while only 40% of companies have official LLM subscriptions — isn't a governance problem. It's a signal. People are finding augmentation value on their own because their employers are still chasing automation.²⁰

5. First-mover advantage is real. When the market corrects — and the data suggests it will — the winners won't be the companies with the biggest models or the most compute. They'll be the ones who built technology that makes humans more capable, not more replaceable.

The Bottom Line

AI 1.0 promised to replace human thinking. It's failing spectacularly, measurably, and at scale.

AI 2.0 promises to elevate human thinking. The data says it works.

The Shift is happening. The only question is whether you'll lead it or be disrupted by it.

Footnotes

¹ Fortune, "Thousands of CEOs just admitted AI had no impact on employment or productivity," February 17, 2026.

² MIT NANDA Initiative, "The GenAI Divide: State of AI in Business 2025," July 2025.

³ National Bureau of Economic Research (NBER), study of 6,000 executives, published February 2026.

⁴ Creati.ai, "Microsoft Copilot Encounters Major Enterprise Adoption Challenges," February 2026.

⁵ S&P Global, as cited in Beam AI analysis of enterprise AI failure rates.

⁶ ManpowerGroup, "2026 Global Talent Barometer," February 2026.

⁷ Deloitte, "The State of AI in the Enterprise, 2026 AI Report."

⁸ Airia, "2026 State of AI" report.

⁹ Wikipedia, "AI Bubble," citing Bank of England warnings and IMF comparisons.

¹⁰ World Economic Forum, "Anatomy of an AI reckoning," January 2026.

¹¹ Fortune, "'We're not in a bubble yet' because only 3 out of 4 conditions are met," February 1, 2026.

¹² ETR, "Enterprise AI Trends 2026: How Leaders Measure ROI and Risk," February 2026.

¹³ TechCrunch, "In 2026, AI will move from hype to pragmatism," January 2, 2026.

¹⁴ Stanford HAI, "A Human-Centered Approach to the AI Revolution."

¹⁵ Stanford SALT Lab, Shao et al., "Future of Work with AI Agents," 2025.

¹⁶ California Management Review, "AI Automation and Augmentation: A Roadmap for Executives," July 2025.

¹⁷ AICorespot, "The Human + Digital Organization — Why 2026 Is the Year of Empowered Workforces," February 2026.

¹⁸ MIT NANDA Initiative, as reported by Virtualization Review, August 2025.

¹⁹ BCG, Wharton, and Deloitte research on AI-human collaboration productivity gains.

²⁰ MIT NANDA Initiative, "The GenAI Divide: State of AI in Business 2025."

Stephen Klein is the Founder & CEO of Curiouser.AI, the Reflective AI company building technology that thinks with you, not for you. He teaches AI Ethics and Entrepreneurship at UC Berkeley.