True transformation takes time
AI's progress is slower than the headlines suggest. True transformation takes years, not hype cycles. Comparing AI's rise to the internet's decade-long march, here's why skeptical insight matters more than blind optimism.
Azeem Azhar's recent Bloomberg piece, AI Will Upend a Basic Assumption About How Companies Are Organized, is one of many articles promising that AI will transform everything. Azhar wrote The Exponential Age, the 2021 Financial Times Best Book of the Year, so he's not some random commentator. But contrast that with Gizmodo reporting that Microsoft is quietly canceling data center leases after promising $80 billion in AI infrastructure. Satya Nadella himself said the world hasn't turned AI hype into "a meaningful lift in the actual economy."
That tension is where we actually are. ChatGPT, Claude, and others keep getting better. But LLMs still make things up. They'll cite papers that don't exist, invent statistics, fabricate quotes. This was a problem in 2023 and it's still a problem now. Maybe it's because they're trained on human content, and humans have gotten pretty comfortable denying basic facts lately.
I use AI every day. I manage an AI team at work. I see what it can do. I also see where it falls apart. On the ground, Nadella's skepticism rings true: companies aren't seeing AI money rain from the sky.
I keep thinking about the web in 1994. It opened up a new world of information, but it took nearly a decade to hit the mainstream. Thirty years later, it's obvious the web transformed the global economy and politics. I suspect AI will do the same, eventually. But for now, I'd ignore the hyperbolic claims coming mostly from academics and media commentators who don't ship products.
Where AI has actually landed
That said, there are real impacts happening. Hospitals use predictive analytics to anticipate patient needs. Supply chains reroute shipments based on real-time disruptions. When DeepSeek released its R1 open-source model earlier this year, the efficiency gains challenged Nvidia's dominance so directly that Nvidia lost $560 billion in market value in a single day. That's not hype. That's a real company losing real money because a better model appeared.
AlphaFold solved protein structure prediction, which had been an open problem for decades. Generative tools produce marketing campaigns and prototypes at scales that would have required entire teams five years ago. These aren't theoretical futures. They're happening.
But as Nadella points out, these disruptions haven't translated into broad economic gains yet. The technology works. The business models are still catching up.
The quieter changes
Most of what I see day-to-day isn't revolution. It's incremental improvement. Netflix recommendations. Google Maps traffic. Salesforce dashboards that surface insights without anyone learning a new system. Small businesses using Shopify's built-in ML to set prices or predict demand spikes.
Journalists and academics focus on grand narratives because grand narratives get attention. But the actual work happening in most companies is boring: automating customer support tickets, managing inventory slightly better, identifying which sales leads are worth pursuing. None of this makes headlines. All of it adds up.
The numbers don't lie
Economist Daron Acemoglu estimates AI will contribute about 0.7% to US GDP growth over the next decade. That's it. Implementation costs and organizational inertia eat most of the gains. Companies adopt AI, then spend years figuring out how to actually use it.
Hallucinations remain a problem. I've watched LLMs confidently cite sources that don't exist, generate plausible-sounding nonsense, and double down when corrected. This undermines trust, and trust is what you need for enterprise adoption. Until that's solved, AI stays in the "useful but unreliable" category for anything that matters.
What comes next
Pharma and telecom will probably lead adoption because they have specific, high-value problems that AI can actually solve. Edge computing will let smaller models run locally, which matters for privacy and latency. Regulation will slow everything down, as it should.
Nadella's skepticism is useful. Not because AI won't matter, but because the timeline is longer than the hype suggests. I use these tools daily. I see their potential and their limits. We're still figuring out how to integrate them in ways that actually help.
The genie is out of the bottle. But electricity took decades to rewire the economy. The internet took a decade to go mainstream. True transformation takes time.
