AI Financial Advice
Discussion
I found this quite an interesting and surprising - and frightening - stat around how many people actually think that what ChatGPT or Gemini or Claude or whatever other AI model tells them is regulated and can be trusted in isolation from any other suggestions or advice.
A third of young investors think chatbot advice covered by FSCS
A third of young investors think chatbot advice covered by FSCS
Puzzles said:
It s mental imo that people blindly follow but it does state its replies with such confidence.
The again, people blindly follow IFAs and they're not guaranteed to be right either... and AI isn't swayed by the thought of fees/commission. But I agree there may be redress from a faulty IFA.Stupid people doing stupid things with their money is nothing new. They're morons, they'll always find new and exciting ways to get past idiotproof.
The problem is that we then get rules to (unsuccessfully) try to protect the morons from themselves which just get in the way of everyone else.
The problem is that we then get rules to (unsuccessfully) try to protect the morons from themselves which just get in the way of everyone else.
JoshSm said:
Stupid people doing stupid things with their money is nothing new. They're morons, they'll always find new and exciting ways to get past idiotproof.
The problem is that we then get rules to (unsuccessfully) try to protect the morons from themselves which just get in the way of everyone else.
I think your spot on, technology just creates new ways to be stupid with money. The problem is that we then get rules to (unsuccessfully) try to protect the morons from themselves which just get in the way of everyone else.
I use AI to do the maths e.g if I pay in £100 per month averaging 8% returns what would it be in 30 years but I wouldn't ask it where to get those 8% returns.
I suppose the AI's could be programmed to tell people to invest in the AI's. Sounds unbelievable but would it really surprise you, these tech bro's don't give a s
t about anyone else. As with most things if you feed it decent data and have a decent understanding of the topic then you can guide and shape it.
It won’t be long before there are models that are approved for financial advice, the likes of Harvey are already used by legal firms in a big way so I’m sure we’re only a matter of time away.
It won’t be long before there are models that are approved for financial advice, the likes of Harvey are already used by legal firms in a big way so I’m sure we’re only a matter of time away.
Jamescrs said:
I don't even trust it to give me the correct tyre pressures for my car, never mind financial advice. I think I'm starting to get old fashioned maybe but I barely use A.I at all, certainly not for anything meaningful
Interesting analogy - I’ve just asked chat for the answer on one of my cars - replied with generically it’s 32- actual answer is 31 so close but no cigar. I find it useful for general enquiries but where you already have a vague idea of what you expect to see as an answer.
alscar said:
Jamescrs said:
I don't even trust it to give me the correct tyre pressures for my car, never mind financial advice. I think I'm starting to get old fashioned maybe but I barely use A.I at all, certainly not for anything meaningful
Interesting analogy - I ve just asked chat for the answer on one of my cars - replied with generically it s 32- actual answer is 31 so close but no cigar. I find it useful for general enquiries but where you already have a vague idea of what you expect to see as an answer.
So far I’ve had AI help with:
1) Understanding how current providers/funds have performed and what their charging structures look like;
2) identifying potential SIPP providers, looking at things such as:
Breadth of investment options
Transfer in/out
Employer contribution flexibility
Hidden charging structure and how to navigate/avoid them.
3) looking through user reviews to identify potential service issues with the shortlisted providers.
4) drafting a series of questions that my current providers are working through in order that I fully understand whether or not there are things baked into those plans that I might not want to lose.
I’m now moving on to fund selection, starting to filter down funds etc to give me the mix and exposures I want (and specifically don’t want or want to know are present). We did a deep dive into one sector in particular to identify companies/funds that have a specific exposure I want that could, in some cases, be secondary to their core business.
It has been tremendously useful *but* that’s coming from a background where I understand what to ask, what AI’s blind spots are likely to be, and how the finance game is played. I might, at some point, try to figure out if there’s a way to let it trade against a small pot of money on a semi-automated basis, but I’m not quite ready for that.
LooneyTunes said:
I m on the opposite side of this one, and have spent a couple of evenings this week using AI to review my pensions - I have a couple and have never actively managed them since they were set up/last consolidated. One plan has been untouched for over 15 years.
So far I ve had AI help with:
1) Understanding how current providers/funds have performed and what their charging structures look like;
2) identifying potential SIPP providers, looking at things such as:
Breadth of investment options
Transfer in/out
Employer contribution flexibility
Hidden charging structure and how to navigate/avoid them.
3) looking through user reviews to identify potential service issues with the shortlisted providers.
4) drafting a series of questions that my current providers are working through in order that I fully understand whether or not there are things baked into those plans that I might not want to lose.
I m now moving on to fund selection, starting to filter down funds etc to give me the mix and exposures I want (and specifically don t want or want to know are present). We did a deep dive into one sector in particular to identify companies/funds that have a specific exposure I want that could, in some cases, be secondary to their core business.
It has been tremendously useful *but* that s coming from a background where I understand what to ask, what AI s blind spots are likely to be, and how the finance game is played. I might, at some point, try to figure out if there s a way to let it trade against a small pot of money on a semi-automated basis, but I m not quite ready for that.
Interesting - albeit with your caveated existing knowledge in mind. So far I ve had AI help with:
1) Understanding how current providers/funds have performed and what their charging structures look like;
2) identifying potential SIPP providers, looking at things such as:
Breadth of investment options
Transfer in/out
Employer contribution flexibility
Hidden charging structure and how to navigate/avoid them.
3) looking through user reviews to identify potential service issues with the shortlisted providers.
4) drafting a series of questions that my current providers are working through in order that I fully understand whether or not there are things baked into those plans that I might not want to lose.
I m now moving on to fund selection, starting to filter down funds etc to give me the mix and exposures I want (and specifically don t want or want to know are present). We did a deep dive into one sector in particular to identify companies/funds that have a specific exposure I want that could, in some cases, be secondary to their core business.
It has been tremendously useful *but* that s coming from a background where I understand what to ask, what AI s blind spots are likely to be, and how the finance game is played. I might, at some point, try to figure out if there s a way to let it trade against a small pot of money on a semi-automated basis, but I m not quite ready for that.
The Fund selection part though would certainly save you a heap of time.
Was that foc Ai or part of a more thorough paid for service ?
I don't think there's any doubt that as augmentation it can come up with some immensely good information - and some immensely inaccurate information.
I've saved a small fortune compared to what I'd have to pay a professional but I've also seen answers where if I'd based a decision off them I'd be completely screwed.
My surprise was more that anyone would think it was regulated in the same way paid for advice usually is.
I've saved a small fortune compared to what I'd have to pay a professional but I've also seen answers where if I'd based a decision off them I'd be completely screwed.
My surprise was more that anyone would think it was regulated in the same way paid for advice usually is.
butchstewie said:
I don't think there's any doubt that as augmentation it can come up with some immensely good information - and some immensely inaccurate information.I've saved a small fortune compared to what I'd have to pay a professional but I've also seen answers where if I'd based a decision off them I'd be completely screwed.My surprise was more that anyone would think it was regulated in the same way paid for advice usually is.
All financial advice is regulated in the UK, so I can see why you might think asking for financial advice would give you either regulated advice or no answer at all though.I have another type of problem that is regulation adjacent and the answers I'm getting are leaning heavily into the legalities around my question rather than giving me a straightforward answer, so it isn't operating in a legal vacuum. It's an interesting state of affairs.alscar said:
Interesting - albeit with your caveated existing knowledge in mind.
The Fund selection part though would certainly save you a heap of time.
Was that foc Ai or part of a more thorough paid for service ?
I have a Claude Max 20x plan, but ran this on Opus 5 instead of Fable (I have another project that is burning that allowance).The Fund selection part though would certainly save you a heap of time.
Was that foc Ai or part of a more thorough paid for service ?
The key is being precise about what you want. The brief I gave it was about one page of A4 long, augmented with the actual fund statements etc downloaded from my existing providers and placed in a shared folder. I asked it to ask questions to refine the research and it came up with 4-5 quite thoughtful ones.
What I did find interesting was that even having completely neglected the plans, they’ve still managed to outperform FTSE by a small amount (even without the fees, which weren’t terrible, being fully optimised). Fee optimisation alone has the potential to make a reasonable contribution to final value.
The part I found most interesting was how the AI tackled my request to dig into the fee models to understand what low headline rates were hiding. The approach it adopted extended as far as digging in to annual reports and the such like to see where/how revenue was being reported.
Good points on this thread, and pertinent for me as I've been playing with AI for stress testing a retirement plan (note, not asking it for investment advice). In summary "Free" copilot, some basic maths on drawdowns, shows I'll be ok based on some simple assumptions, but quite a few errors, not in the maths, just what it thought I wanted, Gemini, all that, plus it chucked in different growth curves for my various investments. Then I tried it on our Enterprise license at work (switched to the ChatGPT 5.6 engine), and it absolutely smashed it, I now have a spreadsheet with graphs showing drawdown rates forecast against various stress test scenarios, amounts taken to maximise tax efficiency and lots of other useful info spread across 7-8 different tabs. All of these options I could have built manually, but the spreadsheet generation takes less than 30 seconds. I have of course verified it for accuracy and it all stacks up.
I did play around asking for investment advice, and it pointed me at gilts/bonds for low risk return, which was interesting and I will investigate further. Using it blindly for anything like that is as pointed out, for the foolhardy.
Also, for what it's worth, for my job (writing IT bid responses), first pass is always AI now (it got good enough about 10 months ago, previously I was better, now I'm not), might take 4-5 refinements, but with a well set agent it's cut my workload by about 90%, and I'm pretty glad I'm retiring soon. My eldest is doing an IT related degree at Uni and I do worry for him getting a job when he graduates.
I did play around asking for investment advice, and it pointed me at gilts/bonds for low risk return, which was interesting and I will investigate further. Using it blindly for anything like that is as pointed out, for the foolhardy.
Also, for what it's worth, for my job (writing IT bid responses), first pass is always AI now (it got good enough about 10 months ago, previously I was better, now I'm not), might take 4-5 refinements, but with a well set agent it's cut my workload by about 90%, and I'm pretty glad I'm retiring soon. My eldest is doing an IT related degree at Uni and I do worry for him getting a job when he graduates.
I've got AI (Bob) to vibe code a pension portfolio manager for me and partner. pulls prices, graphs , etc.
Now added in a sophisticated drawdown engine which calculates how to minimise tax over a drawdown period considering uk tax bands, ISA, SIPP , TFLS tax rules.
Montecarlo simulation + TFLS manager too.
800 bob coins / $400 in inferencing.
Now added in a sophisticated drawdown engine which calculates how to minimise tax over a drawdown period considering uk tax bands, ISA, SIPP , TFLS tax rules.
Montecarlo simulation + TFLS manager too.
800 bob coins / $400 in inferencing.
bennno said:
It s worth being aware that ai gets seeded with information, I d wager some of the investment firms, especially one starting with v, is very actively doing this ..
Which is like SEO being used to sway Google.Am I right in thinking that AI can only trawl the internet looking for answers, and then lashing them together, and if the particular bit of info you want isn't online you won't get it?
Gassing Station | Finance | Top of Page | What's New | My Stuff