The National Bureau of Economic Research asked nearly 6,000 senior executives across the US, UK, Germany and Australia about AI’s impact on their businesses. Around 70% said their companies actively use AI. 90% said it’s had no measurable impact on productivity or employment. And the executives championing it from the stage? They’re personally using it for about 90 minutes a week.
90 Minutes per week, that’s probably less time than the average exec spends doomscrolling on TikTok. It confirms however something a growing number of tech commentators started to say out loud: we’re not in an AI revolution yet. We’re in an AI investment cycle.
Consumer adoption is slow relative to the investment chasing it, and AI the industry has a new playbook. Blame user anxiety. Blame organisational readiness. Blame the environment. That last one peaked in mid-2025 when Sam Altman published a blog post reassuring the world that each ChatGPT query uses roughly one fifteenth of a teaspoon of water. It’s what happens when an industry under scrutiny tries to change the subject rather than answer the question.
But the real answer is simpler and less convenient: AI doesn’t have its killer app yet.
Think about how the internet actually won. Not in theory, in practice. Amazon made it genuinely easier to buy a book over driving to a store. Google made finding information fast, simple and easy. Uber meant you could summon a Toyota Corolla within five minutes, from anywhere in the world. These weren’t marginal improvements. They were things people couldn’t imagine going back on. Once you’d used them, the old way felt broken.
AI doesn’t have that product yet. The closest candidate is coding tools. Tools like Claude Code and GitHub Copilot have real productivity data behind them, developers who use them well are genuinely faster, and unlike most AI use cases the value is immediate and concrete. But coding tools are transformative for a specific group of people. There are roughly 27 million software developers in the world, compare it to the addressable market for a browser or a smartphone. And the argument that easy coding will pull millions of non-coders into building software underestimates how much of the barrier was never the code itself. It was knowing what to build, how to secure it, and who maintains it when something breaks. Most businesses have spent years learning expensive lessons about home-grown software. Slipstream built a business on the idea that technology deployment and support need specialist skills and training many business just don’t want to build in-house. A tool that makes building easier doesn’t change the calculus on whether they should.
Coding tools are the best the industry has right now. They’re the Photoshop of this era: used daily by millions and never touched by billions. Photoshop didn’t turn everyone into a graphic designer. It made great designers faster and gave everyone else just enough capability to produce something that looked almost right but wasn’t. Coding tools are doing the same thing. The skill is still there underneath. The tool just makes it less visible until something goes wrong.
This isn’t a case against AI. It’s a case for being honest about where we actually are.
The internet had a browser. Mobile had the App Store, which turned a phone calling device into an indispensable one. Cloud had AWS, which meant a two-person startup could run infrastructure that a decade earlier would have needed an enterprise budget and a data centre. Each had a specific moment where the abstraction collapsed and ordinary people could feel the value directly, without understanding or caring what was underneath it.
AI is more abstracted than any of them. The value is often invisible: the draft you didn’t have to start from scratch, the answer you didn’t have to search for, the code you didn’t spend 40 minutes on Stack Overflow finding. It’s real, but it’s diffuse. And diffuse value doesn’t drive adoption. It doesn’t make someone feel like they’re missing out.
The killer app is coming. Someone is probably building it right now, and we won’t recognise it until it’s already everywhere. That’s how these things work. Amazon looked like a bookshop, not a logistics company, for years.
But until that moment arrives, the honest thing to say is: the infrastructure is being laid, the hype is running ahead of the reality, and the companies with the most to gain financially are not the most reliable narrators of where we are.
The NBER survey isn’t a story about AI failing. It’s a story about AI waiting.