Britain’s big banks are using artificial intelligence to crack down on people trafficking, personalise customer investment choices and overhaul call centres.
The latest wave of generative AI models are allowing lenders to go beyond traditional machine learning techniques, which have long been used to identify potential cases of fraud and assess credit risks.
Santander has developed an AI model trained to identify suspicious patterns of behaviour in accounts which could point to instances of people trafficking. According to Jas Narang, Santander UK’s chief transformation, data and AI officer, banks have historically been slow to identify this sort of organised crime from customer data. “It has been a little bit hit and miss for all banks in the past,” he says. “And more importantly, it’s not always been timely — it has been analysis [coming after] the event by which time criminals have moved on.”
However last year, the bank built an AI tool which was trained to pick up on certain “tells” that could indicate people trafficking — such as money being deposited into the account from several different locations within a few minutes of each other.
Narang added the difference between traditional machine learning, which is used to analyse vast reams of data, and generative AI, is that the latter can make judgments in a more timely manner. “The difference between what was happening previously and now is the timeliness. It’s picking up stuff whilst criminal activity is being perpetuated. So you can literally pick it up on the day.” Since the rollout of the tool last year, the technology has allowed Santander to generate hundreds of leads indicating trafficking, which the lender then passed on to the authorities for further investigation.
Beyond financial crime, banks are also using generative AI to reshape their consumer-facing services. Lloyds Banking Group is using AI to start personalising the services it offers to its customers. Generative AI bots can scan and analyse data from customer accounts, such as transactions, savings and risk appetites, and use that to provide a personalised service.
Ranil Boteju, chief data and analytics officer at the UK’s biggest high street bank, says the aim is to bring the sort of bespoke financial advice the ultra-high net worth individuals receive to millions of users.
This month, Lloyds announced that it was tasking 7,000 staff members with training up an automated financial assistant which could give advice on managing finances to clients.
The ‘agentic AI’ assistant — the term given to models which are set up to behave autonomously — is being tested and is expected to be rolled out next year. Customers will be able to discuss certain payments with the financial assistant, but it will not be advising on any type of regulated activity.
Boteju says Lloyds hopes to build on the model, allowing users to customise their preferences so that the assistant would then be able to act on their behalf with a series of “personalised nudges”, such as by putting savings into an Isa. “In the future, savings could be automatically invested into Isas if customers have agreed to it beforehand — which could help them better save for their future,” he says. “So, it will provide the guidance, and then the difference with our agentic AI assistant is it will then start to take action on the customers’ behalf.”
Scott Marcar, NatWest’s chief information officer, sees a similar pattern in how the technology will evolve: “Today [AI] brings greater speed, personalisation and protection from online threats; tomorrow it will power even more seamless and hyper-personalised experiences as we advance next-generation AI capabilities to anticipate customer needs faster and more effectively than before.”
Generative AI has also transformed the due diligence involved in the lending process. While banks have long used machine learning software to analyse reams of financial data they already hold, generative AI can pull information from multiple different formats.
One example is the due diligence required in lending to a commercial property client. “Before providing a loan, lots of bank statements, reports, collateral and a whole host of other documents have to reviewed,” says Boteju. “You end up with 10 or 15 different documents all in different formats. The real estate lender has to piece through those, look at them and because they’re so un-standardised, traditional robotic process automation would never work, because everything’s different. But with generative AI, you can now automate that so it extracts all the key information, puts it in a simple format — and so rather than the person having to do that over a course of an hour, it takes a couple of minutes.”
Not all banks are using AI solely to enhance services. The adoption of AI-powered customer service bots has allowed digital bank Klarna to cut its workforce in half in recent years through natural attrition, although it has had to reverse tack to an extent after some unsatisfactory results using the tools. Still, two-thirds of Klarna’s customer service operations are now automated.
Boteju says Lloyds is not using the technology to cut jobs. The bank developed an AI tool which brought together all the lender’s information in one place, which sped up the search time for call centre staff by 66 per cent.
Marcar, from NatWest, also says AI has allowed the group to “cut complexity” and “free up colleagues to focus on what matters most for customers”.
Narang, at Santander, says while the applications of AI are vast, organisations need to be careful about developing tools experimentally or pursuing any “pet projects,” adding: “It’s a very, very clear cut business case up front in terms of either customer benefit and or productivity benefit.”
Read the full article here