Probably.
Look, statistics is a weird subject if you haven't spent much time with it. A lot of it is counterintuitive to our day to day lives, and most people simply don't understand it (and I strongly put myself in that bucket).
And as a result, the same goes for predictive forecasting.
This is one of my more unpopular opinions, but for most businesses it's simply not suitable as a core forecasting methodology.
Sure, there are some amazing tools out there for this, machine learning, AI, etc etc. But outside of very expensive (and involved to set up) enterprise tools, most accessible processes for predictions rely on taking simple historical data and running simple processes on it.
Most businesses simply change too much over time for that to work. You're introducing products, services, locations all the time, and the ones you were doing last year, you're changing too.
You can disagree with me if you like, but I offer this advice: if you want to put this functionality in budgetholders' hands, first do this: don't, but alongside the usual budget, run the predictive models, then compare the variances over time.
If it turns out the predictions are better than what the bottom up process comes up with, great, send me a message; I'll buy you a coffee to apologise for being wrong.
But my experience is that they're either hilariously unsuitable, or to make them vaguely suitable you simply spend as much time tweaking them as you would building up budgets the usual way. With the gigantic added risk that budgetholders just push the button and blindly trust the results, without using their knowledge to correct them.
A fantastic illustration of the illusion of statistics comes from Spurious Correlations - here's my current favourite (apologies if your marriage ended over Disney movies):
