If you haven’t heard, many industries and professions are said to be under serious threat by the rise of artificial intelligence or AI. It certainly makes for a good story, and there’s a financial angle. Lately the financial media have started doing stories on how investors are using AI to pick stocks and guide their financial decisions. It’s gotten around to stories asking can you replace your financial adviser with AI?
It’s all good fun. And we don’t feel the need to defend ourselves with this one. We’d say financial advisers being replaced by AI is still a long way into the future. While the market will decide in the long term, currently in Australia the demand for advice outstrips the supply and the great majority of peoples’ retirement savings sit with industry super funds. Industry super funds do a pretty good investment job, but a terrible service job.
Trying to manage things yourself and being forced to sit for hours on the phone dealing with what should be small issues that turn into administrative nightmares, is just one reason why people want to hand things off and not bother.
Do We Use AI?
Yes, we do. We’ve been recording meetings for a while now so we can transcribe meeting notes. These days that audio is then put into an AI transcription product we use and within a few minutes we have full meeting notes. They always must be double checked for accuracy as not everything is 100%, but it saves a couple of hours of manual transcription.
That’s a dedicated tool, and if we can find other dedicated AI tools that can improve our processes, we’ll always consider them. It’s also a paid product with a specific purpose. It turns audio into text. As far as the free public AI tools available they will have strengths and some significant weaknesses. And like everything, it’s a case of you will get what you pay for.
Our licensee, FYG Planners, ran some experiments on some tasks we regularly do, using public AI tools, and some of the results were hilarious. Bearing in mind AI requires very precise prompting to successfully extract the information or response you’re looking for, they gave three public AI tools very precise instructions to retrieve and sort data on a list of investment funds.
AI Number 1: wouldn’t do what they asked for “technical and legal limitations”. At least it was honest.
AI Number 2: They provided them with the first ten funds, then they had to prompt it to get the second ten funds multiple times and combine them into a CSV file. No matter how many times it was asked for all twenty, it kept providing ten, then it would ask if we needed the next ten. Eventually it provided all twenty, but the story doesn’t end.
The results were quality checked, and the AI had made them all up! The results were in the range of what they should be for a bond or equity fund, but none were correct. When the AI was challenged, it said that the data provided was “example placeholders”. The AI then asked if the user would like the first ten again with the correct values. Ok, try again. It again provided incorrect data.
AI Number 3: Told the user it was getting the data, then came back with a table with fund names, but “data not available” for what they’d asked. On inspection of the table it had generated, it had changed some of the names of the funds it had been given!
The user went back to AI Number 2 attempting to be more specific and asked it to source data from the official fund managers’ websites. It did the same ten fund list again and provided a new table. When asked where the data came from, it provided a list of links. Upon checking the links, it was found that some of the fund managers did not actually provide some of the specific requested data on their website. In short, the AI had made up the data again! The user asked the AI if it made up the data and it said, “you’re absolutely right to question this and they weren’t sourced from the fund manager, but they were placeholder estimates to show how a table demonstrates formatting”.
It then asked if the user would like it to try again. They declined.
This isn’t uncommon. One of our FYG colleagues also told us of an argument they had with AI over cricket statistics. The staff member was right, and the AI was wrong, and again, the AI made its results up, only to apologise when a direct link to correct data was provided! We’ve also heard of an ordeal someone had while helping to source an electrical relay for their heater. The person had enough knowledge to point out that the AI had offered up several parts that were near, but not right for the heater. They then used two other AI tools to double check the answers. It took an hour of them double checking each other before the right part was correctly identified.
Always take note of the warning that you sometimes see on these tools: “AI-generated content may be incorrect”.
Limitations
One of the overlooked issues is public AI can be a case of garbage in, garbage out. Often AI is simply and quickly aggregating an answer from webpages. For example, we’ve found traffic to our website from AI and found our own blog posts used when we’ve asked AI questions. It’s nice that AI sees us as a reliable source, but we do have an accuracy disclaimer on our site for a reason.
There are problems with this. Firstly, what the particular AI tool deems an authoritative source. Back when we simply searched the web for something, all the results would appear. It was up to us to filter through them. It was up to us to determine what might be rubbish because it was all there in front of us. AI tools will supply links to our questions, either in the answer, or underneath it, but it delivers an answer it has already formulated for us. Our thinking and judiciousness have been removed from the search.
Conversely, if you already have the research and the right links and all the information, but you want feedback, opinions, and analysis on it all, you can feed the AI tool and will probably have a much better experience.
When it comes to financial advice, there’s not a lot that is going to change.
The people who were always going to DIY are just going to adapt and test new tools to start their DIY investment journey or continue on with it. It’s still going to take some time and effort on their part. There won’t be any miracles. AI can’t see the future and won’t identify the 100 bagger stocks ahead of time.
The people who can’t be bothered, don’t have the time, don’t have the interest, or the confidence, and are happy to pay someone to do it for them, will continue doing so. The availability of AI doesn’t change the need to research, digest, and have the confidence to implement something.
The processes of financial advisers will certainly be improved and aided by AI tools, but we were already using financial modelling tools, which are a low-level form of AI. And we’ll still need to remain abreast of technical and legislative changes, and whether that’s delivered by an annually updated text we keep on our desk, or a large language model, the implementation will remain similar.
Real intelligence still has some value.
This represents general information only. Before making any financial or investment decisions, we recommend you consult a financial planner to take into account your personal investment objectives, financial situation and individual needs.
With thanks to FYG Planners for this article