We were wrong before about AI killing jobs. Here’s the proof.

In 2016, a very smart, very informed fintech CEO wrote a guest commentary for Technical.ly. He confidently predicted that the number of financial advisers in the US would be cut in half within a decade.

“There’s simply no reason to pay for something when a higher-quality product is available for free,” he wrote, pointing to the rise of robo-advising. At the time, there were roughly 200,000 financial advisers nationwide.

Today? There are more than 250,000, per the Bureau of Labor Statistics.

So instead of a 50% drop, we saw a 25% increase.

To be clear: his prediction about the technology was spot on. Robo-advisors, launched in the late 2000s, were a very real early use case for machine learning in consumer finance. By 2016, they were growing fast — and they still are.

But the prediction about jobs? Dead wrong.

That’s the lesson. Rightly predicting a technology’s rise does not necessarily correlate with rightly predicting its effects on employment. It’s the heart of a bet I have with a friend: Will there be more software developers working in May 2026 than there were in May 2025? This has real effects for economic development strategy. 

It helps to understand the ways several of us got that bet wrong a decade ago. Why did financial advisers grow even as the automated systems really did advance? In part, because we humans are messy. A 2023 YouGov survey showed that more than half of Americans didn’t even know robo-advising existed. About a quarter outright refused to use it. Many others had mixed feelings.

Meanwhile, the technology didn’t eliminate advisers — it shifted what they do. By automating basic tasks, robo-advising lowered the cost of entry for clients and freed up advisers to focus on more complex financial planning. Tech-forward financial planning firms such as eMoney Advisor and startup Wealthmore changed more than replaced financial advising. 

This isn’t unusual. In computer science, there’s a concept called “the overhang.” It describes the gap between what’s technically possible and what people actually want to do, for cultural, social or regulatory reasons. Just because a software engineer can theoretically replace financial advisers — or retail clerks with self-checkout — doesn’t mean customers want this. 

Sometimes, we just don’t want the most efficient thing.

We could be seeing it again. An influential MIT paper last fall argued the real economic effects of AI’s rollout will be far more muted: Just 5% of jobs in the American economy could be meaningfully automated away in the coming years. To be clear, that’s still millions of positions, but far from a jobs apocalypse. 

What of recent changes to tech hiring? Well, at least as much of that appears to be caused by a pandemic over-hiring bonanza and rising interest rates. A recent Financial Times analysis argued that tech hiring appears to be picking back up again.

We’ve been here before. And we’ll be here again.

When we talk about artificial intelligence and its potential to “replace” jobs, it’s worth remembering: We can be right about how powerful a technology is — and still wrong about how we’ll actually use it.