Are we starting to see an expectation reset from the AI industry leaders? AI expectation vs reality are coming back together.
In quite a dramatic turn, AI CEOs went from saying LLMs would wipe out 50% of entry-level office work to saying, “my intuitions [about the impact on entry-level white-collar jobs] were just off.”
What’s striking is that the reset is coming from every side of the table: the people building AI, the people financing it, and the people buying it.
The Builders
As a reminder, Dario Amodei (Anthropic) was saying AI could soon eliminate 50% of entry-level office jobs (Axios, May 2025). Amodei still believes an impact on jobs is forthcoming. But OpenAI CEO Sam Altman has changed his tune.
Sam Altman (OpenAI CEO), speaking recently at a Commonwealth Bank of Australia conference in Sydney:
“I don’t think we’re going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about.”
“I thought there was going to be more impact on entry-level white-collar jobs being eliminated by now than has actually happened… But that is an area where my intuitions were just off.”
The Investors
There is a change from investors as well. David Solomon (Goldman Sachs CEO) argued in a New York Times piece that 25% of work hours may be automated, not 25% of jobs.
“While A.I. eliminates jobs in some sectors, it may lead to job growth in others.”
The Buyers
Large customers of AI compute are also starting to see a more realistic scale of benefits from AI, particularly in coding, which was the first industry to embrace AI. Andrew Macdonald (Uber’s COO) says that while they are producing a lot of code, they aren’t seeing an increase in problems solved (Rapid Response interview):
“You make your head explode, right? When you hear companies talking about, hey, 25% of code commits over the last quarter were AI driven… our token usage went from X to Y…”
“That link is not there yet… It’s very hard to draw a line between one of those stats and okay now we’re actually producing like 25% more useful consumer features.”
Conclusion: The real bottleneck is deciding what is worth working on.
AI is a magical tool. If you described what it could do to technologists 10 years ago, it would blow their mind.
It was easy to make the connection that if white-collar workers were automating a large chunk of their tasks, then companies would only need a few super-efficient workers and the rest would be laid off. But it seems early users are automating their low-value tasks so they can spend the extra time on high-value ones. The bottleneck was never producing the work; it was deciding what’s worth doing.
Thanks to the Primeagen for this idea and pointing me to the quotes. Watch his video here: Maybe we were wrong
Colin Finkle is an industrial designer in who has integrated AI into his design workflow from the earliest of days. He is from Burlington, Ontario Canada.