For big stuff like portfolio allocations? Hell no. It's a solved problem with a few good solutions already. Broad market cap weighted indexing, target date funds. Any custom themed portfolio the ai comes up with is bonkers. It may chase recent past performance and load you up with a DRAM themed portfolio. Which may be great for a few years but then you get left behind when you fail to rotate out of dram at the right time. Cap weights have rotation baked in.

For small stuff? Maybe. If it can give some insight into total cost of owner ship. on things like cars. Used vs new. The ai could compare price charts, depreciation, etc to rank some good value buys. But that can be calculated with traditional non-ai logic, just needs data.

The easy part in providing the model with enough context about your existing portfolio etc.. the challenge is about understanding your goals and optimizing towards them in the long run in terms of expenses (ongoing and then things like house, saving for kids), investing (portfolio mix, asset allocation, access to alternatives) and tax strategies. A model can do a good job on each one of those vectors separately. Whether it can do so holistically on all vs. a human professional (who's good at his job + has access to models) I think is a much harder question. This ofc becomes much more true if you are dealing with a complex portfolio, or longer time horizons. Hope this adds something to the convo.

LLMs? No. 85% correctness and avcuracy — even 95% — is just not going to cut it in that context.

Not with my money. No way.