From AML to personalised products: how AI is transforming financial services

The World Economic Forum (WEF) has added its voice to the chorus of caution about the rise of artificial intelligence (AI) technologies, specifically in financial services. “The long-term impacts of AI may be even more radical and transformative than we first imagined,” said the Swiss strategists. “The very fabric of the financial services ecosystem has entered a period of reorganisation, catalysed in large part by the capabilities and requirements of AI.”

As it happens, those words were written five years ago, in August 2018, when the WEF published a detailed report, ‘The New Physics of Financial Services: How artificial intelligence is transforming the financial ecosystem’. The research, in collaboration with Deloitte, involved over 200 interviews and six international workshop sessions over the course of almost a year.

Half-a-decade on, the impact of AI appears to be less radical than predicted. However, IOB associate faculty member John Curry, who is also founder and chief data scientist at Last Mile AI, a boutique AI consultancy that works with banks, fintechs and other organisations, says there is more going on behind the scenes than people realise.

In the beginning: machine learning for fraud detection

AI in the financial services sector has come through the ‘hype cycle’ of inflated expectations and disillusion to begin to deliver gains, according to Curry. A key development was the evolution of supervised machine learning (ML), where computers are ‘trained’ with large volumes to data to recognise what is normal or abnormal and set parameters.

“Data is the fuel for all of these systems,” Curry says. “Banks using AI are in a great position because they own these incredibly rich data sets. AI, applied to data analytics, can do things like detect fraud in real time and flag particular transactions.”

On an individual level, people will be familiar with instantly receiving a message from their bank if they use their credit card in an untypical location or for a larger amount than usual. Curry says gains are being made in the more sophisticated use of AI in anti-fraud and anti-money laundering measures.

During the critical know-your-customer process of onboarding a new bank customer, for example, AI can be applied to augment the traditional ‘knowledge graph’ approach and look at a customer’s relationships to other individuals and entities. This can include shared company directorships, family connections and other relevant links.

“Machine learning has become more powerful and can deal with much larger data sets and more heterogeneous data sets – for example, images and graphics as well as text,” says Curry. “It can help detect potentially problematic information for opening an account or extending credit.”

Powering up underwriting

AI has been quietly “powering up” actuarial models at financial institutions for many years, according to Curry, allowing banks to automate or semi-automate some processes, such as loan applications. “AI tools can pull together data sources and come up with a [credit] score,” he says.

The value of AI in financial services is that it not only rapidly processes huge amounts of written or transaction data, but can also recognise and understand patterns between data. For now, most banks still have hybrid models, with some human involvement in decision-making.

The trend is only going one way: in a 2021 survey of almost 500 financial institutions in nine countries by IT services group NTT Data, more than 80% said they were already incorporating AI into their strategies. Even more, 83%, said AI was a “critical part” of their strategy to attract and retain customers, expand into new markets and run more profitably.

Applying AI to credit underwriting has benefits for both lenders and consumers, according to a 2022 report, ‘AI for Good: Research insights from financial services’, from the Washington-based Center on Regulation and Markets at Brookings. However, like Curry, it cautions that the outcomes from AI are only ever as good as the data that goes in.

Your own personal bank

Within machine learning, Curry pinpoints ‘deep learning’ - which teaches computers to learn by example - as a critical technology. Real-world examples include apps such as Spotify, Netflix or Amazon offering recommendations based on your history.

Many banks, fintechs and insurers already use AI for basic personalisation of products and services; for example, an automated message about a new loan or insurance quote when an existing product is up for renewal. The NTT Data report says banks must up their game; it identifies a cohort called ‘Futurists’, customers aged around 35-44 who are willing to share more data with financial institutions in return for more personalised services.

“For banks to succeed in the future, they need to provide hyper-personalised, positive experiences that form an emotional connection,” says NTT Data.

Curry sees banks forming partnerships across industries - with retailers, telecoms operators and healthcare providers, for example – to offer AI-powered services such as hyperlocal recommendations. A customer considering an electronics purchase could be geolocated to a specific store, identified with facial recognition and approved for extra credit in real time.

In the insurance sector, an internet-connected car can send driving data back to the insurer, which can inform an underwriting decision and lead to a lower (or higher) premium for the driver. “What’s coming down the tracks are personalised products where you can pick the bits that are relevant and jettison the bits you don’t want,” says Curry.

Trading on technology

It is a basic rule of business that investment goes where the highest returns are likely to be made. For bigger banks and asset managers, that means applying AI and ML to trading activity; according to study by JP Morgan in 2020, over 60% of trades over $10 million were executed using algorithms.

Foreign exchange trading is dominated by AI algorithms, which can trade based on price movements around the globe and around the clock. Algorithms remove emotion and biases from trading and use predictive analytics to respond to market signals within microseconds.

“Forecasting tools are getting more sophisticated,” Curry says. “You see ML-based forecasting more and more in options pricing and portfolio forecasting optimisation.”

The arrival of AI-based trading tools in the hands of retail investors, meanwhile, is breaking down the traditional power structures between providers and punters. As a result, financial services organisations are “facing customers who are armed with personalised AI for customised trading” and consider themselves more sophisticated than in the past, says Curry.

He cautions, however, that AI applications based on large language models such as ChatGPT and Bard can generate incorrect or misinterpreted results and are poor on analysis. “They also tend to be overconfident and can be arming consumers with the wrong information,” he says.

Policing financial services artificial intelligence

The continuing evolution of AI in financial services necessitates a balancing act between better services and protection of consumers and their data.

Existing GDPR legislation around data and privacy is helpful says Curry, and he expects that an upcoming EU AI Act is likely to be “fairly stringent”. The European Council aims to reach agreement with EU countries by the end of this year on the final form of the Act, which intends to ensure AI systems used in the EU are safe and respect existing laws.

For its part, the Central Bank recently published a research report on Data Ethics Within Insurance, many aspects of which are applicable to the wider financial services sector. It considers the uses, benefits and risks associated with what it calls ‘Big Data & Related Technologies’ in the sector, including AI.

For now, a WEF comment from 2018 still has some resonance: “The public discourse on AI in financial services is highly sensationalised,” it said, “creating an excess of both exuberance and fear.” Where the situation stands in five more years will be interesting to see.

Boost your digital skills

Ciaran Fennessy is Module Coordinator for IOB's Professional Diploma in Data Analytics in Financial Services. Are you looking to strengthen your expertise in digital and innovation? Explore IOB’s full portfolio of digital and innovation programmes to develop specialist knowledge and skills that will equip you to succeed in the ongoing transformation in financial services.