Sam Sivarajan is a keynote speaker, independent wealth-management consultant and author of books on investing and decision-making. His forthcoming book will explore how to thrive in a world of uncertainty.
What is happening with AI market volatility? After years of breathless headlines and surging stock prices, tech leaders are sounding more grounded. Infosys Ltd. INFY-N chair Nandan Nilekani joined Microsoft Corp.‘s MSFT-Q Satya Nadella and IBM Corp.‘s IBM-N Arvind Krishna in suggesting that we’ve entered the “trough of disillusionment” – the sobering dip that often follows the initial excitement over new technology.
Stocks such as Nvidia Corp. NVDA-Q, Palantir Technologies Inc. PLTR-Q and C3.ai Inc. AI-N have captivated investors, at least until recently. But the real question isn’t what artificial intelligence can do – it’s whether we’re ready to trust it. “For the first time, we intend to place trust in non-human intelligence for decision-making. And that leap of faith is not an easy one,” Mr. Nilekani said.
That line gets to the core of the issue. The limits holding AI back right now aren’t primarily technical – they’re human. Behavioural economists call this reluctance the ambiguity effect: People tend to pull back when information is uncertain or incomplete.
We’ve seen this before. In the late 1990s, investors chased internet stocks well past their fundamentals. It took years before real value emerged. The same might be happening with AI. Goldman Sachs says tech firms are set to spend more than US$1-trillion on AI in the coming years. But so far, this spending has produced more flashy headlines than hard results. As MIT economist Daron Acemoglu has warned, meaningful change will take time, maybe even decades.
It’s not the first time a big idea got ahead of itself. In 2003, people hailed the completion of the Human Genome Project as the start of a health care revolution, but the transformation isn’t coming nearly as fast as promised. Cold fusion made a similar splash in the 1980s. It’s not that these technologies failed – it’s that the timelines were longer and more complex than investors expected.
Nvidia’s recent stock pullback – down nearly 30 per cent from its high – might indicate that we’ve hit “peak AI.” The mood is shifting. Investors are no longer just buying the story. They want evidence. How is AI improving performance? Where are the returns?
And this is where the paradox gets sharper. The places where AI could be most helpful – health care, finance, government – are exactly where trust is hardest to build. A consumer AI application can afford to get your dinner recipe wrong. But a financial algorithm making credit decisions, or a diagnostic tool analyzing a biopsy, can’t afford to hallucinate.
A recent McKinsey report found that while AI adoption jumped sharply – from 33 per cent of firms in its survey using generative AI in 2023 to 71 per cent in 2024 – more than 80 per cent of respondents reported no measurable earnings impact. An MIT Sloan study echoed this finding: 75 per cent of firms are experimenting with generative AI, but only 3 per cent say they’ve integrated it at scale.
The barriers? Not the tools. The people. Companies reported issues with trust, resistance to change, and the challenge of integrating new systems into old workflows.
After initial enthusiasm, employees retreated mostly to the sidelines. MIT found that 59 per cent said AI tools hadn’t changed how they work at all. Among those who did use them, most spent less than an hour a day doing so.
We also tend to misjudge how to use these systems. MIT research found that AI-human teams often underperform compared with the best AI or best human alone – unless we redesign the workflow so each plays to its strengths. AI thrives at pattern recognition. People excel at creativity, judgment and navigating grey areas. Getting the mix right takes effort, time – and trust.
That trust is fragile. Algorithm aversion is real: Studies by Berkeley Dietvorst and colleagues show people quickly abandon algorithms after a mistake, even when those algorithms are more accurate than human judgment. But allowing people even minimal control or input makes them far more likely to stick with the system. Trust increases with transparency and agency.
For investors, this presents both risk and opportunity. The most valuable AI companies will be the ones that build trust systems around their tech – offering clear explanations, dependable outcomes and ways for people to stay in the loop. That, in many ways, is the antithesis of the “black boxes” we are currently seeing.
Health care is an excellent case study. AI could transform diagnostics. But adoption remains slow, because neither doctors nor patients will rely on a system they don’t understand. It will be the firms that build explainable, transparent AI – helping doctors make better decisions, not replacing them – that will position themselves for long-term success.
Three takeaways for investors: First, they should prioritize companies that are building trust infrastructure around AI – including transparency, explainability and human-AI collaboration – rather than focusing solely on technological sophistication or, worse, marketing hype.
Second: The real value from AI will emerge gradually, through practical, incremental improvements, not sudden breakthroughs – so long-term thinking is essential.
Third: Successful AI adoption isn’t simply about what the technology can do, but about how well organizations enable people to trust and use it effectively.
Productivity gains from AI will come when organizations figure out how to embed these tools into human workflows and earn people’s trust.
As Warren Buffett has said, “Be fearful when others are greedy, and be greedy when others are fearful.” The same applies here. It’s easy to get caught up in either hype or cynicism. But the real opportunity lies between the two: in clear-eyed assessments, grounded implementation, and a focus on the systems that make new tools work.
In the end, the most important breakthrough isn’t AI itself – it’s building the trust and integrations to use it well.