Kamala Harris’s big loss on Tuesday night caught almost everyone off guard. Despite widespread expectations that she’d be at least slightly ahead going into the election, the reality turned out starkly different: she got crushed.

In those critical battleground states—Pennsylvania, Wisconsin, Michigan, Arizona, Nevada—where many assumed she had an edge, Trump surged past expectations.

Just days before the election, the Des Moines Register poll, one of the most respected in political circles, had Harris leading by 3 points in Iowa. The New York Times and Siena College polling also showed her ahead in several battlegrounds, with Trump solidly up only in Georgia and Arizona. But these numbers were way off on election day.

Even in typically blue strongholds, the polling was off. In Maryland, where Democrats usually don’t even blink at the polls, Harris underperformed her polling average by over a percentage point, while her Republican opponent exceeded expectations by 4 points.

Even in New Jersey, another traditionally blue state, polls were wildly off the mark. Rutgers ran a poll in mid-October that missed Trump’s numbers by double digits, and even the most accurate polling underestimated the gap between the two candidates by six points. 

But this isn’t the first time polls have missed the mark by such a wide margin. It happened in 2016, too, when pollsters underestimated the support for Trump because their traditional methods didn’t reach the “silent” Trump supporters—those less likely to take a survey call or respond to pollsters. The same trend seems to have repeated itself in 2024, raising the question: are polling methods outdated?

It’s clear that something needs to change and perhaps artificial intelligence may be the answer. Traditional polls rely on people actually picking up the phone and answering questions, but AI could do so much more. 

By analyzing enormous amounts of data in real time—everything from shifts in demographics to social media sentiment—AI has the potential to capture a far more nuanced picture of voter sentiment. 

This shift could mean fewer reliance on who answers a call and more focus on where people’s attitudes and thoughts are actually trending.

One guy who didn’t get it wrong in 2016 or in 2024 is my guest on Wealth Formula Podcast this week: Jim Richards. Jim has a unique perspective on why polls keep getting things wrong, even as voter behavior changes and political dynamics shift. On this week’s show, we discuss that as well as his new book on how artificial intelligence will affect the economy and national security.