Balancing AI efficiency with human expertise in asset management

You can also listen to this podcast on iono.fm here.

SIMON BROWN: I’m talking with Andile Jonas. He is the head of marketing at Momentum Savings. Andile, appreciate the time. In a recent note out from you, you make the bold claim that you are ‘an enthusiastic AI user’. You summarise reports, you even soften criticism in work emails. That one I’m going to start using.

But money is a place where you draw a line and you say, ‘Thanks, but not with my money’.

ANDILE JONAS: Indeed. I think the advent of artificial intelligence has done wonders for our ability to learn, our ability to automate some routine tasks. It has been absolutely incredible.

But I think the role of the financial advisor is well and truly still safe in the hands of the experts. I think AI can enhance that role, but not quite yet replace it, Simon.

Read:
Why AI is not a financial planner
Here’s why you need a financial advisor IRL
The rise of AI in financial services: Balancing innovation and integrity

SIMON BROWN: I get your point. There are large language models and they really are just sort of piercing pieces of words together. They’re not necessarily intelligent. They certainly don’t necessarily have the breadth of knowledge, they don’t perhaps have the experience.

They also perhaps don’t have the intimate knowledge of the person to whom they would be consulting. They actually fail in a number of areas in terms of financial advice.

ANDILE JONAS: Yes. Context is everything, isn’t it? I think there’s something really important in the process of giving advice, where your advisor understands your relationship with your family, your relationship with money, your relationship with the future, your dreams.

All of that context shapes the advice that he or she gives you. And some of those things you may not even be aware you’re telling or showing your advisor. AI doesn’t know to ask those questions. So context is everything.

SIMON BROWN: Yes, it’s actually a great point. It doesn’t hurt to ask them.

You talk around the ‘cognitive debt’ and how that fits. Talk me through the concept here around this, and how this also potentially trips things up.

ANDILE JONAS: Yes. I think I’ll perhaps allude to it by saying AI doesn’t know to ask. There’s something quite important in the process of doing the homework, reading about the options that you have, that helps you understand a little bit better what you want.

There’s the critical thinking that you go through in yourself when you learn the difference between a 30-day call account and a retirement annuity. That helps you understand what you want out of your money as well.

Read: The future of retirement planning: How AI is changing the game for South Africans

I think the absence of that work, that two or three steps that you take yourself, leaves you with a cognitive debt where [from] the solution presented to you by this AI engine called ChatGPT or Claude, you can’t tell whether or not this is appropriate or not because you just haven’t done enough of the of the reading, of the understanding of what the thing is.

I think AI is wonderful in explaining  complicated and complex things in simple language. At the very least, we should do that. Otherwise we’re left with this thing called ‘cognitive debt’, where I can’t tell if this is a good or a bad answer.

SIMON BROWN: That’s a great point. Also, they’re trained to be agreeable. I use Claude and I’ve tried to tell Claude, ‘Stop being nice to me, tell me when I’m wrong’. And Claude kind of struggles.

But I can see a world in which markets are down, I don’t know, 20%, 30%. You ask your LLM [Large Language Model] if you should sell. And LLM is going to say, sure. It just doesn’t know. It’s like, ‘Well, that seems like a fun thing to do right now’.

ANDILE JONAS: Yes, the ‘tyranny of politeness’. I think you’re absolutely right. There are sort of clever prompts to try and take that point away, but you made the point that it’s trained to do that. It’s trained.

One of the things I fear most about it is that it always sounds confident in its answer, even when confidence is not the right sentiment at that moment.

So a certain degree of scepticism and doubt is warranted on occasion, particularly when there’s money involved, where deeper thinking, deeper homework, deeper investigation is needed.

I think to answer a question about which share I should buy with the greatest of confidence can sometimes be misleading to John Q public.

Read: AI trading tools can hallucinate and outright lie

SIMON BROWN: Yes. I’ve done it. If I say to it, ‘Give me, I don’t know, the hating on it option’, it does it. And then it goes all in on that and that comes back to you.

It’s what you alluded in the previous answer as well. So much of financial advice these days is behavioural coaching, is talking me down from the ledge, is stopping me doing silly things and the like.

That’s one area where, again, AI is just not going to know me well enough. It’s not going to know the circumstances. It’s not going to have what I think is a huge part these days of an advisor’s job.

ANDILE JONAS: Hundred percent. I think AI should automate some of the admin, not the relationship. There’s a value in understanding somebody’s appetite for risk. But not just that, because that’s the first question.

I’ll tell you a story. My financial advisor, one of the first questions they asked me was: ‘If you’re not around – let’s assume just for research purposes that you’re not around and you pass away tomorrow –  who in your family should I be worried about? Who should I be most concerned about for two reasons? One is to look after them, or two, to be concerned about their ulterior motives.’

Now, thankfully I didn’t have this problem, but I loved the question. I absolutely loved the question. I think it shapes how you put your structures together.

That’s just not a conversation I would ever be able to have with a large language model [LLM].

Read:
When ethics and evidence outweigh automation
Autonomous investing: The future or just an AI fad?
How AI will transform the investment industry

SIMON BROWN: Hundred percent –  a conversation you could have it with it. I know a lot of doctors are using it. I’ve a friend who has experienced this in the US, where the LLM is recording the conversation for notetaking, freeing up the doctor to sort of have a proper conversation. There is perhaps space for it, just the right space.

ANDILE JONAS: Hundred percent. Look, if you’re not using AI to create efficiencies and capacities in your day -to-day, then you’re missing a trick as well.

So there’s definitely an opportunity here to get scale, to get speed, to get efficiency using AI – and it’s here to stay.

So the more acquainted with it and the better we are with it, playing with it, we’re going to get sharp.

But I think to completely replace that which a human being brings to a conversation and a thinking might just be a step too far.

People talk about making investment decisions. You’re actually making a life decision when you’re talking about your money and your future.

SIMON BROWN: Yes, that’s actually a great point. It’s way more than just investment.

And the last point  – we haven’t even touched on it – of course, there’s a regulator here. There’s the FSCA. Momentum will be an FSP, have a licence there. I just don’t know whether LLM fits in that. And frankly, I don’t think anyone – probably including the LLM – knows either. That’s a huge gaping hole and a regulatory nightmare.

Read:
The advice gap: Why young South Africans stopped asking professionals what to do
Information Regulator calls for responsible AI adoption by businesses
Bank CEOs’ AI obsession collides with warning from watchdogs

ANDILE JONAS: Yes, it is, it is. I think you’re right. I think not enough of that conversation is happening at the moment, because if you make some really, really bad financial decisions on the back of AI advice, what do you have to fall back on? Who has done that research, who has done the homework of finding out why the fund choice was what it was?

I think that’s the scary part – that people may, in fact, find out a little too late. So I’m a big, big fan of what AI brings to the party, but just a good old-fashioned expert is the right person to give you advice.

SIMON BROWN: Yes, I agree, 100%. And someone who has that FSP has the regulation and truthfully does what it does. It means that I do my homework, I’ll make sure I know the product, I’ll make sure I know my client. And I think the behavioural part, that coaching part, is so important.

But we’ll leave it there. Andile Jonas, head of marketing at Momentum Savings, appreciate the time.

#Balancing #efficiency #human #expertise #asset #management