The use of AI for financial guidance has risen sharply in recent years. In the United Kingdom, around 28 million adults used artificial intelligence tools for personal finance advice last year, according to Lloyds Banking Group.
However, experts caution that AI-generated advice may not always be practical. Danilo McGarry, an AI adviser to the Chartered Institute of Personnel and Development, told Financial Times, that when an user asked a chatbot to help in reducing their tax bill, the AI bot suggested moving to Monaco,
Though this is technically correct, the advice was of little use to someone living and working in Croydon, highlighting how AI can sometimes overlook real-world circumstances when offering financial guidance.
This week, a study by Fidelity International found AI use to be particularly concentrated among retail investors aged 18 to 34, with 36% of this age group using the technology to support their investment choices, compared with 29% of 35 to 54-year-olds and 5% of over-55s, FT reported.
AI can be useful, but has limitations
AI can make financial information more accessible, especially for people who may not have easy access to professional advice. But experts warn that it should be used with caution, as relying on inaccurate or misleading can lead to costly and time-consuming mistakes.
Above all, experts warn that users should remember that AI is a tool, not an authority, and is thus prone to error. Specialists in AI ethics warn that the tech is more susceptible to mistakes than many people realise, FT reported.
Why giving a good prompt is the key
Much of a AI chatbot’s ability to give the most accurate response depends on the user’s prompt, hence being specific in your requests is also essential.
While chatbots are increasingly good at prompting you for additional information, asking AI how to invest the cash in your savings account, or what products to invest in, without providing details about your income, tax position, time horizon or risk appetite, is likely to generate answers that may not cater to your financial requirements.
McGarry told the news publication that he has witnessed the potential consequences of giving vague in real time. “I have seen someone paste a portfolio screenshot and ask, ‘should I rebalance?’, without telling the model their time horizon, risk tolerance, or that half the holdings sit inside a Sipp,” he recalled.
“The AI happily told them to sell, which would have triggered an unnecessary tax event in the taxable portion,” he added. This happened because AI does have reasoning skills like humans do. Instead, it generates plausible language based on patterns in the data sets used to train it.
According to Horowitz, this can lead to outputs that may be technically reasonable in isolation, but fall apart once a real-world context is taken into consideration, FT noted in the news report.
Details shared with an AI chatbot are not private
Experts also caution users to think carefully before sharing sensitive financial information with . Many people assume these conversations are private, but information entered into AI tools may be stored, reviewed or used to improve future versions of the technology.
As a result, users should avoid sharing confidential details such as bank account numbers, passwords, tax identification numbers or other personal financial data.
According to the news report, Farrell-Kingsley warns that free AI tools tend to have a “trade-off”, the price being an individual’s personal data. While working on LLMs in the past, she reports having seen highly sensitive user input, including their card and banking details.
So before giving AI your personal data, an user must consider whether they would be comfortable if their data were no longer entirely private. If the answer is no, it’s worth reconsidering how much information you disclose.