Three things ‘experts’ got wrong about AI

Everyone's a pundit. That would be fine, except that everyone is wrong, especially on AI and the economy.

I've found myself getting annoyed hearing everyone's definitive proclamations on this still-nascent market. Why does it feel like people are so confident about an industry with no precedent, little available data, and inherently limited history from which anyone can claim expertise?

Consider these three things I heard about a year ago — many confident proclamations from folks in the industry and none of which have turned out to be true.

1. AI will be commoditized

The price would come down, models would converge, and spending would level off as the price of tokens went to zero. Instead, average monthly AI token spend on Ramp has grown 13x since January 2025, dwarfing every other business spend category we track.

2. AI will create a hiring boom for software engineers

The Bureau of Labor Statistics still projects a nearly 20% increase in hiring for software engineers over the next decade. Instead, employment at software companies fell in 2024. The latest data from private sector datasets shows a similar slowdown in tech hiring.

3. Only model companies will capture value

Remember the "AI wrapper" discourse? Cursor went from $1 million to $2 billion in annual revenue in 26 months, built entirely on foundational model APIs. It's one of the fastest-growing companies according to spend data from Ramp.

What the data actually says

This is an extraordinarily uncertain market, yet it attracts so many definitive claims from actors without access to data on actual spending behavior. The so-called experts often have a profit motive. Those who don't tend to hedge so much they say nothing useful at all.

The purpose of our Spring 2026 Business Spending Report is to give you more certainty. We track over $100 billion in annual business spend across 50,000 companies. That's the dataset behind all the data in our report. We're confident about what we know well, and we make no misrepresentations.