Opening sesame

Digital financial assistants are coming to a bank near you, but developing bots that know the right thing to say at the right time and that can run all your financial affairs for you will be much harder than some suggest, reports Ouida Taaffe

Taaffe-(digital-assistants)-smartphone-genie.jpg

We’ve all heard the story of Aladdin, in which a genie summoned from a lamp grants every wish. The popularity of smartphones arguably rests on offering a similar sense of control and power. One touch and it’s done.

But smartphone ‘genies’ with far greater ‘magical’ capabilities may not be far off. At the start of December 2024, for example, Google announced a step forward in attaining its “vision of a universal assistant”, with a new research prototype that can “understand and reason across information in your browser screen, including… text, code, images and forms, and then uses that information via an experimental Chrome extension to complete tasks for you”. Or as a future marketing campaign might have it: simsalabim.

A universal assistant is still a work in progress. What we are likely to see first is the arrival of digital financial assistants: less powerful than Google’s ‘universal assistant’ but still a fundamental shift. What does that mean for sectors such as retail banking? Can there be a single ‘genie’ that, once summoned, effortlessly optimises people’s finances for them?

There are companies that would like to be able to offer one. In a June 2024 interview on the Acquired podcast, the Chief Executive of Klarna, Sebastian Siemiatkowski, said: “In a few years, you will see out of all of this emerge three or four really large global players in the space who are basically being that digital financial assistant of customers. That will be a very valuable position to have. That is basically what we’re trying to do.”

Klarna wouldn’t discuss its plans, as it’s in a quiet period in the run-up to an IPO, but it’s not the only company that sees change coming. Banks expect it too. “There is a huge opportunity for digital assistants to be a leapfrog technology in getting better outcomes for consumers,” says Miles Hillier, Product and Technology Director, NatWest. “We can see how artificial intelligence could help provide simpler, more personalised and more convenient financial services.”

What would good look like?

But what would a good digital financial assistant look like? “It would be like having a real financial assistant, just behind your shoulder, always just chirping in your ear at the right time,” says Andrew Stevens, Principal, Banking and Financial Services at Quadient, a provider of customer experience technology (among other services). “For example, it could say ‘You know, you’re considering going on holiday. This is the budget that I think would be OK for you rather than the £1,000 you have left in your bank account’. Banks could do that very well.”

But if people turn to their digital assistant on their Android or Apple phone to manage their finances, could banks find themselves ‘disintermediated’ – relegated to being the ‘invisible’ providers of regulated credit and payments services? Most people use smartphones and increasingly few of us use bank branches, particularly the young. In the UK, around 98% of those aged between 16 and 24 have a smartphone, according to Ofcom.

There is a huge opportunity for digital assistants to be a leapfrog technology in getting better outcomes for consumers

Smartphones also collect a lot of data, which is why they can support so many helpful applications such as GPS navigation apps. Just as importantly, consumers can opt to share their banking data with selected third-party service providers, via open banking. Open finance, when it comes, will only expand the amount of financial data – including data on mortgages, pensions, investments and savings – that can be made available to third-party service providers. The government’s end goal is ‘open data’ in which consumers can share data from other ‘key sectors’, that is, as things stand: water, healthcare, retail, energy and telecoms providers. Data from Big Tech will be part of this, if regulators deem it useful in providing financial services.

Nikhil Rathi, Chief Executive at the Financial Conduct Authority (FCA), said: “Big Tech’s growing emergence in financial services has already made life easier for consumers, but it is still unclear how valuable their data will become in financial markets.” If the FCA does find that data from Big Tech should be included, “it will look to incentivise more data sharing between Big Tech and financial firms through its Open Banking and broader Open Finance work”.

Digital financial assistants work better on simple transactional things. When it becomes more complicated, that’s when a human being is often required

As things stand, smartphone data may not be much of a game-changer. "Is data from smartphone operating systems a useful proxy for anything finance-related?" says Stevens. "Only right on the peripherals. So, for example, when someone's charging their phone, they're unlikely to have it in their pocket. That means it's a bad time to be trying to give someone advice. The data that sits in an operating system is useful, but it's useful in terms of how you communicate with them."

The data used in large language models (LLMs) that power the AI systems that consumers are familiar with may also not be that useful. "Today, LLMs have around 95% of publicly available data in them," says Prashant Jajodia Managing Partner, Financial Services Sector Leader, IBM Consulting. "If you think about enterprise data, there is less than 1% of that in LLMs... Models trained on bank data will be much more efficient."

Jajodia also points out that "good design and implementation are all about ensuring that data privacy is maintained. Sensitive data should only be available to those authorised to have it".

Hillier at NatWest argues that banks are "in a slightly differnet position to other industries. We're able to provide a digital financial assistant to deliver a more personalised service without relying on third parties".

He says that banks already provide insight on managing money, for example in goal setting. “In the future, you could see a digital assistant going beyond that to help with budgeting and saving. In the near-term, we aim to help customers self-serve rather than provide regulated advice via the app. However, AI could be used in the future to help our bank colleagues who are qualified to provide advice to offer even more personalised experience, help and support.”

Klarna is interested in launching digital financial assistants partly because, as a firm, it has access to so-called ‘level 3’ data. That is, Klarna knows not just that a certain amount of money was spent where, but also what it was spent on. It can see, for example, that someone bought cream puffs and a treadmill and, in principle at least, use that to extrapolate further purchasing and financing needs.

Do banks need access to so-called ‘level 3’ data to be able offer digital financial assistants? “The most important thing when it comes to data is transparency over what is being used when, and how,” says Hillier. “We already have access to lots of valuable data that we can use in thoughtful ways.”

Stevens sees level 3 data from merchants as “dangerous”. “A digital financial assistant should be there to take care of your financial well-being,” he says. “In an AI world, level 3 data allows someone to understand what a person has spent money on and would be used to then recommend other products. It would encourage spending, which would be great for a company that has a vested interest in more spending, but a digital financial assistant should prioritise the right thing to do.”

Taaffe-(digital-financial-assistants)-AI-and-feeli

Using AI to help bank customers navigate fraud scenarios is one of the most complex retail finance use cases for the technology

Stevens points out that the use of level 3 data can also be frustrating for consumers. “The fact that I bought a broom from Amazon doesn’t mean that I want to see adverts for five more brooms. I’m not collecting brooms.”

He says that a digital financial assistant tends to work better for simple transactional things. “Where it becomes more complicated, that’s where a human being is often required at the moment. I’m not sure where, as an example, I would explain to an AI bot the importance of my relationship with HSBC. The fact that, yes, I can get a better rate elsewhere, but I worked for them for 17 years so that’s not all that matters to me. I understand the quality of service that I get. I understand the fact that when I have a problem, I will be dealt with incredibly carefully and sensitively. That stuff is missing from a digital financial system.”

Hillier says of NatWest: “We have over 300 years of heritage and 19m customers. This is something we’re not complacent about. Our focus is to continue to deliver the best digital and colleague-based experience in the market.”

It’s all in the complex mix

Data that is not obviously about finance can, however, be useful in providing an excellent customer experience. Ian McKenna, Chief Executive of AdviserSoftware.com, says that some financial advice firms in the US use technology to profile customers psychologically to discover their learning style. “If somebody is a visual learner and you send them a 40-page text document, they ain’t reading it,” he says. “These firms work to align advisers with clients so that they have the same learning style… Some of these systems will actually take the content from a text-based scenario and create the visual content for you. They can do that in a fraction of the time that it would take a human to do it.”

That ‘know your customer’ aspect of data analysis can also be used to help protect the customer, but it takes more than smartphone data. The GSMA, the global association of mobile operators, and UK Finance, a banking and financial services association, announced at the end of November 2024 that mobile network operators and banks in the UK, including NatWest, have collaborated on ‘Scam Signal’, which uses real-time network data to identify correlations between phone calls and authorised push payment scams.

Overarching data analysis is not easy to do. “Using AI to help bank customers navigate fraud scenarios is one of the most complex retail finance use cases for the technology,” says Hillier. “AI, machine learning and a variety of advanced data techniques are already being used behind the scenes to spot trends and activity that could be dubious, to help better support our customers.”

Could Big Tech bring together all of this disparate data from many different apps across different sectors, and from the underlying networks themselves, in real-time, to provide a digital genie?

Some technologists have strikingly big ambitions. One of the co-founders of online bank Monzo, Jonas Templestein, recently argued that AI agents could devise, build and run new firms, including raising the capital. “No humans required,” he says. “This would require major reforms in the legal and financial system.”

Ouida Taaffe is the Editor of Financial World