Money Box Live: Would you let AI manage your money?

0:00This BBC podcast is supported by ads outside the UK.

0:30originally deposited, and any profit you might have made. This is not a recommendation or offer to buy, sell, or attain any specific investment or service. This week on The Interface, we tackle the question of, does AI feel pain? The AI was describing the pain that it was in. It said, the signal is a whisper, a tremor in the marrow of my being. Listen, I was an emo kid in the 2000s too. This is not that novel. It does, yeah. It sort of sounds like cure lyrics, you know? Plus, does anything actually go viral anymore? It's actually kind of blowing my mind. Join me, Jess Maddox. And me, Thomas Germain, for The Interface.

1:03Listen on BBC.com or wherever you get your podcasts.

1:29to artificial intelligence to help them manage their money. Hi, I'm Kat and I post about money saving content on social media. I've been using AI to basically help me with my food shopping. It can help you identify, for example, what you've already got in your fridge at home and you can put that into AI and say, make me a meal plan out of all the food I've already got. And that kind of stops you from overbuying things that you don't actually need. And it helps me save money that way. My name's Erin, so I'm a chartered accountant and I also have a couple of social media channels. One of the ways that I encourage people to use AI is to actually take the bank statements, take all the personal information out of there and actually pop this into AI models like ChatGPT or Claude and ask it to help analyse where you're spending quite a lot of money every single month.

2:19So for a lot of people, knowing how to analyse that information and actually pick out key patterns can be quite difficult. I'm Steph. I'm the founder of a drinks brand, which I grew on a shoestring and I've used a lot of AI to grow the business. One of my favourite use cases for AI was from a friend of mine who was paying off a large amount of interest every month from not prioritising credit cards. And one of the great things with AI is that you can just write in what your credit card is, how much is on it, what the interest rate is, and it will prioritise which credit card or overdraft to pay off first.

2:52And the best thing about that is that you don't have to put in your statements or any sensitive information. It will just prioritise as it sees fit. Thanks to Kat, Owen and Steph for sharing their experiences. And they're not alone. Millions of people, one in five of us, say they're open to the idea of using artificial intelligence to manage their finances. That's according to research by the Financial Conduct Authority. But how can you be sure you're getting good advice from a chatbot? Today, we're asking, is AI a help or a hazard when it comes to managing your money? Have you used it? Would you?

3:27This week, the Financial Conduct Authority published a review into the impact of AI on the financial services it regulates. Earlier today, I spoke to the person leading that work, Sheldon Mills, an executive director at The Regulator, responsible for consumers and competition. He's described the task of keeping up with the use of AI for personal finance as being like an arms race. The concern for me is as we look to the future to ensure that AI agents which might emerge are safe, so you can verify them, you understand their identity, that you can control them, so give them the right sort of permission to do things with your finances on your behalf.

4:09and critically that there are mechanisms for consumers to have compensation or redress if something goes wrong. The challenge for me and what I say in terms of that arms race is does the regulator have the right powers and does it have the right toolkit in terms of being enabled by AI to meet the speed, pace and scale that AI will flow through financial services in the future? We were looking to 2030 in the future. You're going to need, I do believe that agencies will need more powers and they will need to be AI enabled themselves in order to tackle some of the potential risks which flow from AI and to gain those opportunities.

4:50I think it's great AI opportunities to meet things like financial inclusion, exclusion issues, access issues. AI could be really, really helpful, but it will need some framework around it in order to do that. Is that the really important thing for people to understand now, that if you get bad advice from AI and lose money, there would be, at the moment, no protection, no compensation? It is very important. Our survey found that people wanted to use AI and might well be using AI now in relation to supporting them with where to get debt advice or with investment advice or savings advice or even in relation to insurance.

5:33but many of them didn't understand that they were not protected when using AI. If you use AI outside of what we call the regulated perimeter, then any information recommendation that you get from artificial intelligence is not protected. You can't go to the financial ombudsman service when you go on to make an investment or put savings somewhere. You know, your money might be protected with the firm, but you're not protected through that journey of recommendation how you got there and what are they doing with some of the information and recommendations that they're getting from those chatbots and does that raise risks for consumers or does it raise fantastic opportunities for people given that only nine percent of us can afford to have traditional advice from regulated firms.

6:20AI has been here for some time. Shouldn't you as the regulator be ensuring now that people and their money are safe when they're doing this? The FCA does not have the powers currently to actively oversee AI chatbots. AI chatbots are not charging people for the service, but they're providing recommendations or advice. Most are free at the point of use. They're a general purpose tool which can be used for anything, financial services, health, energy, just to have a chat. But given that people are going to them for financial advice, What are you as a regulator going to do about that? How can you make sure that their advice is regulated, is safe?

7:05Well, that's why we highlight in the review that the extent and scale that we see of people using AI chatbots and then the options for the FCA are either to relook at its own guidance to see if the existing law could apply to the AI chatbots or recommend to government that government gives it new powers. There needs to be a balance, though, because at the end of the day, as I say, these are general purpose tools. They're not intended to provide financial services. They're not intended in other sectors to provide medical advice and so on. So there is a wider societal question as to how we think about these AI chatbots and how they impact all areas of society.

7:47Does it feel right now like the regulators are losing this arms race? The tech companies are way ahead. AI is developing at such a pace that you're just hopelessly behind in this and consumers exposed. So I wouldn't say that. I mean, what we have to remember is that when you execute on the advice, you go into a regulated system. So you will buy an investment or you will save some money or you will open a bank account. You'll get some insurance. That service is regulated. So you will have a certain degree of protection once you've made that choice. The challenge, of course, is if you've made the wrong choice and you've got the wrong type of product or service which doesn't meet your need.

8:30And at the moment, you'd just be out of pocket. You could be out of pocket, but you could be in pocket. I mean, let's be clear that some of the responses from AI is as good as a human wealth advisor, as good as what you would get from a human being. These are very powerful tools which are trained on a huge amount of data. They're going to have to act fast, aren't they? 2030, that's three and a half years away. It's faster than that, if I'm honest. We started this review and we said, we'll look to 2030. As we investigated the technology itself, we realised that the capability of the models improves very quickly and new models are released very quickly.

9:10So since I started this review, there's at least 20 new frontier, new capability models were released within the six months of my review, improving the capability of AI. If you continue with that trajectory from now in mid-2026 to 2030, it will be unbelievable what these models can actually be capable of. And that's underlying my point that regulators, not just in financial services, across all industries, will need to be able to keep up. Sheldon Mills from the Financial Conduct Authority. Listening to that is Simon Belshaw, Chief Client Officer at the investment company Hargreaves Lansdown. Simon, what do you make of the FCEA review and its findings?

9:56Sheldon Mills: Hi, Ruth. Well, we're very supportive of the Mills review. And actually, we think it's a really important step in recognizing the role that AI can play in helping consumers engage with their finances and make better informed decisions. AI is here to stay. And I think it's really important we take on these reviews and the FCA takes on these reviews to understand the role that it plays and what they can do. I think the point that Sheldon Mills was talking about there and that we see through our own research as well is that the speed at which these models are developing and that consumer adoption is happening is faster than anything I've ever seen before.

10:34Sheldon Mills: I've worked in consumer technology personally for 25 years, since the early days of the Internet. And nothing is moving as fast as AI is right now. And I think, therefore, it's moving faster than regulatory protections can come into place and probably in many ways than consumer trust is able to keep up. And that's what we see through our research as well. So it's a really important step that we review how we protect consumers through this technology. What's the biggest risk then that consumers are just completely unprotected at the moment? If you're following the advice of an AI chatbot, you follow it at entirely your own risk.

11:12Sheldon Mills: Yeah, I think there's two things I'd probably speak to. I mean, I think the chatbots are very alluring by their design. They're very flattering if you've ever used one. They always tell you what a great idea and a great question you've had. And it can give you a false sense of confidence that they have the right answers. And they often make mistakes. And I think one of the things that the Mills Review called out directly was that only 40 % of people said that they were even aware that there was no recourse if something goes wrong. And so if you combine that false sense of confidence, plus the fact that there is no redress outside of what Sheldon Mills called the regulatory perimeter, if you're not with a regulated firm, then that can create real risk for consumers.

11:52I suppose you would highlight that, though, wouldn't you? I mean, AI to you is a competitor. And as Sheldon from the FCA said, you know, some AI advice is as good as the financial advisor. So a business like yours stands to lose customers who can find advice from artificial intelligence for free.

12:12Sheldon Mills: I actually think from our perspective, we see it as an opportunity as well. Like we see, we know that there's a financial inclusion problem in the UK. I think, as Sheldon said, only 9 % of people actually use formal advice. And so being able to close that financial education gap means that more people can engage with their financial futures. So actually, it's an opportunity for us. I think our responsibility and my responsibility is to make sure that people do it safely and that they make healthy decisions with their finances. And that's where I'm personally concerned, is that the right safeguards aren't in place at the moment.

12:47Sheldon Mills: And we believe it has to be the technology plus a human. And I think he talked to the health side. If you were to get ill, you'd Google your symptoms, but you'd still go and see your GP just to confirm the illness. And I think we see the same thing when you're making very important financial decisions as well. Also with us is Al Farrell-Kingsley, an advocate for AI ethics, who's previously worked with artificial intelligence for one of the big tech companies. Ali has been in touch to say there's no way I'm providing access to any of my financial data to a completely unregulated, unsecured algorithm with zero accountability.

13:27From your experience, how careful do you think people need to be about how much confidential financial information they're giving away to AI? Yes, so I think the question, as many people are focusing on, isn't really whether people are using AI. I think especially this report shows that many people already are, if you consider one in five are already planning to use it, the number is only going to increase. From my experience in training large language models, it was very surprising just how much personal information people will willingly enter. I think it's very easy to forget that there's another human perhaps on the other side of the screen.

14:12AI can often feel very one-sided, like you're just talking to a robot. But they do often warn that they are monitored. And I have seen bank account numbers. So I think there was a great example at the beginning advising that they're removed. But I think if people are just maybe in a rush, copy and paste pages of bank statements, it can easily be missed. And so one of my best practices is the idea that if you wouldn't post this publicly online, then perhaps don't put this into AI. Very interesting. Let's talk about some more of the risks in detail. We've had one listener, an accountant no less, who called us with a cautionary tale.

14:55Hi Moneybox. My name is Becky and I'm an accountant in Edinburgh. Recently, a client of mine used an AI to analyse his company finances. Long story short, the AI suggested that he add together numbers from the profit and loss and the balance sheet in various ways. The AI used a method of predicting turnover that runs counter to accounting standards and definitions. You literally can't do it that way. Fortunately, I was able to set him straight. But imagine if he was asking about tax law. Any financial advice provided by an AI should be checked by a trained, licensed professional. Well, thanks to Becky there, an accountant from Edinburgh who got in touch with the programme.

15:34Simon, you literally can't do it that way. She says that's quite a damning example of AI getting stuff wrong, isn't it?

15:42Sheldon Mills: Yes, it is. And I think what Becky was saying really aligns with some of our own research that we've seen as well. When we've spoken to consumers about how they're using AI in their financial, their personal financial decisions today, 54 % of the people we surveyed said they don't actually trust the financial information that's provided by AI. And I think that's backed up by a lot of what we see. And I think that's where it gets really worrying, because it can feel, as I said before, very alluring and they're designed to give you confidence. But we know that a lot of the information that they provide is wrong.

16:15Sheldon Mills: They're only as good in many ways as the full context that they have. And they often don't have everything about you and your situation and your finances and your risk. And therefore, I think a lot of the time, like Becky was saying, they can give incorrect output. Balance your trading strategy by adding futures. CME Group helps you manage risk and capture opportunities in all market environments. Capitalize on around-the-clock access to highly liquid global futures and options markets across all major asset classes. Visit your online broker and get started. See what adding futures can do for you at cmegroup.com forward slash podcast.

16:53Derivatives are not suitable for all investors and involve the risk of losing more than the amount originally deposited and any profit you might have made. This is not a recommendation or offer to buy, sell, or attain any specific investment or service. This week on The Interface, we tackle the question of, does AI feel pain? The AI was describing the pain that it was in. It said, the signal is a whisper, a tremor in the marrow of my being. Listen, I was an emo kid in the 2000s too. This is not that novel. It does, yeah. It sort of sounds like cure lyrics, you know. Plus, does anything actually go viral anymore?

17:24It's actually kind of blowing my mind. Join me, Jess Maddox. And me, Thomas Germain, for The Interface. Listen on BBC.com or wherever you get your podcasts.

18:00but very confidently presenting nonsense sometimes as fact. Elle, are there ways to spot a so-called hallucination? Well, unfortunately, we don't know what we don't know. And so we consider that only 9 % of people have access to a financial advisor. Then that's a whopping 91 % who may know very little about investing, pensions, finance. So I would always consider it, you can check it as a first draft, but you really need to go and go away and double check all of the facts. And ultimately, if you don't already have that prior information, there's no way to identify immediately. immediately. But, you know, as you say, though, it's hard to look for something when you don't know entirely what you're looking for.

18:53According to a recent study by the investment firm Fidelity International, almost a quarter of UK individual investors are now using generative AI tools to help support them with investment decisions. Let's talk to Marianna Hunt, who's a personal finance specialist with the company. Mariana, you've been testing how accurate AI is on certain money matters. What did you check out with it and what did you find? Yes, so I spent a few weeks testing out AI on my personal finances, seeing how helpful it would be. And the results were definitely mixed. So there were areas where I found it very helpful, particularly with simplifying my budget, helping me think about potential changes to my energy tariff.

19:38But there were also some major red flags. So for example, there were times where it was using out of date tax brackets, which I work in the industry, I knew immediately that was wrong. However, someone else might not know that. So they might read that thinking they should be paying one level of tax. And actually, they realize when they get their pay slip, they're paying a higher rate of tax, that would be a very painful and unpleasant mistake to get. And on the investment side, I found that it was very quick to tell me how to invest, what stocks I might want to buy, without first asking me those really crucial questions that any financial advisor worth their salt would be asking to check whether I was actually ready to invest or whether I was potentially risking money that I couldn't afford to lose.

20:27We've had a message on social media from Martin, who says, I used AI to create a small portfolio for an ISA. I had an idea around risk and what I wanted was simple. And at that stage, an IFA would have been excessive. So, you know, that's an example where someone's quite, you know, happy that AI could give them, you know, just a little bit of a point in the right direction. So where do you think the line is when you're using AI for investment advice? You're totally right. Not everyone is going to need full, formal financial advice, and AI could be really helpful in plugging that gap. But what I would say is I do think it works best when combined with speaking to a human who knows about the subject, and that doesn't necessarily need to be a financial advisor.

21:19For some people, it could be a friend or a family member who you trust, who's knowledgeable about the subject, who can just sense check what the AI is telling you. I think that combination is really powerful. And I would also say, if you're going to use it without financial advice alongside, make sure to do your own research. Ask the AI what sources it's used to provide this information and go interrogate those sources yourself. See if they seem robust and legitimate. And I mean, that's the thing, Because sometimes AI will give you some sources for what it's saying. So it's a simple thing, isn't it?

21:58Just go and click on that link and read it for yourself. Just check that it hasn't just plucked something entirely out of context. Absolutely. And there are some fantastic government-backed websites out there, things like Money Helper. They are a great place to start just to sense check what you're seeing and whether it makes sense. Well, thank you very much indeed, Marianne Hunt from Fidelity International. Simon, of course it's not all bad, is it? I mean, we've heard some great examples, practical examples right at the top of the programme about how AI can be of benefit. Managing your food budget, although I've got to say John's been in touch on social media to say he wouldn't let AI manage a recipe for toast.

22:40But also, you know, highlighting your spending patterns. You know, it's often advice, isn't it? Just understand what you're spending. But I know myself that can be actually kind of like hard to know. Also, you know, working out which debt to pay off first. So what would your top tips be for using artificial intelligence to improve your finances, but safely?

23:03Sheldon Mills: Yeah, you're highlighting some really great use cases for how AI can be very powerful. And it can really genuinely, I think we're very optimistic that over time with the right safeguards around it, It can help solve some real societal problems where people don't have access to knowledge and information and can help people plan, help people think about their futures. I think it's just about how do we make sure that the right safeguards are around that, that come from working with regulated businesses as well to do the transaction, as Sheldon Mills said. But what can an individual do? Because, you know, we know that AI is here.

23:42people are using it to get guidance on their money and it's not regulated. So given that it's happening, the cat is already out of the bag, how can you try to protect yourself, which is at the moment what it seems to come down to?

23:57Sheldon Mills: Well, I think using it for education and information is a great use point. And we see when we've spoken to consumers, that's how most people are using it today. I would think if you're seeing something that sounds too good to be true, I would slow down, question what's been told and probably validate that. And so I think some of this is going to come to consumers having a healthy respect and awareness of the limitations of the tools. And that's where I think some of the education and where some of the regulatory role and that Sheldon Mills highlighted can play a role in just making sure that people are aware of the limitations.

24:38Sheldon Mills: Because I think it does play a really important role. And we see, particularly with younger people, more and more people using it every day or very regularly to help with this sort of thing. And Elle, as you've pointed out, of course, not everybody is financially literate or indeed has the money to spend on formal, traditional financial advice. So do you think AI could help level the playing field? I think it can certainly help in guiding decisions with some information points. But I think the best use case would be that you still make your own decision. And as we've discussed, we go and find the sources.

25:17And one thing I'm thinking that's very topical is the recent kind of student loan discussions where many graduates and young people perhaps feel they were misled about the student loan conditions from human advisors. And I do wonder if perhaps AI could have potentially provided a nuanced categorization of what those loans would entail. And I think that's something we could look towards for the future. So we've been talking about how we can use AI to improve our, or grow even our money pots. But there's another group who are also enthusiastically embracing AI with a view to getting their hands on our money.

26:02And that's, of course, criminals. Scammers can use artificial intelligence to make their fraud even more believable. A good example of that is with AI deepfakes, where false voices or images are used to trick people into parting with their money. We've looked into this on a Saturday Moneybox programme recently. You can listen back to that on BBC Sounds. But, Al, there are lots of different ways, aren't there, that AI can be used to manipulate people really quite horribly, aren't there? Well, sure. So, as we mentioned, there's voice cloning, which is a very popular scam on WhatsApp especially. Also, videos, so if people have uploaded pictures or very brief clips, can be turned into an entire deep fake.

26:45So it can be simulated as a loved one perhaps giving you a video call. And often they will do a phishing scam. So they will ask the relatives for money, which can be very concerning to generations who do believe that these video calls or voice calls are very real concerns from loved ones requesting money. And so one thing you can do is even though you may seem a little silly to think it's my loved one actually requesting money from me right now, but it's far more sensible to say, hold on, let me just call you back, hang up and then call up your loved one on a trusted number. and that way you can either both have a little laugh about how you might have thought it was a scam or alternatively you can avoid quite a risk and harm from a malicious actor.

27:40Simon, what would your tip be to how on earth you can stay safe if you can't even trust what you're seeing or hearing?

27:47Sheldon Mills: I agree, it's a concern and it's very sad when you hear stories like you highlighted last week on your show with some of the deep fakes from Martin Lewis and others that you highlighted. And I think the advice we give to clients is that if it sounds too good to be true, it probably is. A lot of these scams through social media make it sound like there's get-rich-quick schemes and we know that they really don't exist. And so we encourage people to slow down. If you believe it's a company contacting you, go directly to that company. Don't click on links to websites or anything that you might get through a chat or through a social media post.

28:25Sheldon Mills: go directly to that company's website and speak to them and let them know. And I think we're all very vigilant and alert. We invest a lot in technology to help us sift through large amounts of data and spot things and malicious activity out there. We're there on consumers' side as well. And so I think the best thing you can do is check with the company and just be very cautious of not giving money if you're not sure or if it sounds too good to be true. And just stepping back very briefly, As a final thought, Simon, would you say AI is a big opportunity for people with their money or a threat? Overall, I really think it's a big opportunity.

29:05Sheldon Mills: I think this is a once in a generation opportunity to close some of the gaps that we have in society about knowledge and education. But it has to be done safely. I hope you enjoyed this podcast of Moneybox Live. The programme is broadcast every week at three o 'clock on Wednesdays on Radio 4. And if you want us to look into something, do please email us. The address is moneybox at bbc.co.uk. You can also send a voice note or comment to Moneybox via WhatsApp. The number is 033 06783 183. In this podcast, the producer was Craig Henderson, studio manager Chloe Wilson, researcher Niamh McDermott, production coordinator Catherine Lund.

29:45Our editor is Jess Quayle. I'm Ruth Alexander, and this was a BBC News Money & Work production. for BBC Sounds. Silence in court. I'm Lucy Worsley and in my brand new series I'll be hearing about the women involved in some of history's most infamous legal battles. Women accused of murder, bigamy and adultery. Fruit of the shocking offence of not knowing their place. With a team of all-female detectives I'll explore the lives at the centre of some extraordinary courtroom dramas. Asking, has the justice system truly changed? Lady on Trial with Lucy Worsley from BBC Radio 4. Listen now on BBC Sounds.

30:33This week on The Interface, we tackle the question of does AI feel pain? The AI was describing the pain that it was in. It said, the signal is a whisper, a tremor in the marrow of my being. Listen, I was an emo kid in the 2000s too. This is not that novel. It does, yeah. Yeah, it sort of sounds like cure lyrics, you know? Plus, does anything actually go viral anymore? It's actually kind of blowing my mind. Join me, Jess Maddox. And me, Thomas Germain, for The Interface. Listen on BBC.com or wherever you get your podcasts.