Why is there no personal finance software that use back-data with A.I. or regression modeling to make suggestions or to build plans?
What would they be trying to predict?
The value YNAB and Mint provide is objective truth about what you've spent. They can force you to think about the tradeoffs inherent in budgeting by showing that you've overspent one category, and making you decide where to find the money to cover it. They can call your attention to a credit card swipe that's larger than you intended, to a subscription you didn't intend to keep, etc. by just generally getting you to read and think about your transaction history and the sums of transactions per category and overall. Prediction doesn't really enter into it.
One way to understand Mint's business model is as a service that collects training data for machine learning models that do try to predict things, such as how stock prices will move or whether users will click on certain ads.
How would they make money from it?
They sell you the software for $100 (US example; could as easily be 100 Euros or 10,000 Japanese Yen). You use it to make recommendations on your blog. Your blog becomes rich from advertising. They sold $100 worth of software. If they spent $1 million in labor developing it, they're way behind.
Another problem is that the software would stop working and need adjusted periodically. This is easy to do on a server but annoying on a PC. And who pays for the adjustments?
Put both those things together, and it's a lot easier to do on a server. Another advantage is that a server can get a better data feed as well. Pay a premium for the detailed information rather than relying on public sources. And people are used to renting server access where they expect to buy software once.
Another issue is that they are unlikely to beat the market this way. Yes, AIs have done so. But that's the latest AI, constantly adjusted. This is going to be a previous generation AI. It's more likely to match the market. And we already have a way to match the market: an index fund.
If someone had a brilliant AI, the best use would probably be to sell it to a fund manager. The fund manager could then use the AI to find opportunities for its existing investors. Note that a $10 billion fund with a 10% return that gives a .1% commission would be paying $1 million. And that has no marketing or packaging overhead. Think $10 billion is a lot? Fidelity has $2 trillion.
Consumer facing finance is heavily regulated. You are liable for the recommendations you make; if they are based on a black box you risk problems when sued. It is difficult to explain in a court of law why a neural network came to a particular conclusion.
It is much easier to provide advice (models) in the "educated counterparty" market.
Not only do institutional investors in general expect to pay for a quality advice (consumers in general expect to get online advice for free) but the legal implications are different.