Banks look beyond digitisation post-pandemic

Much has been written about digitisation among retail banks, digitally-driven challenger banks and even big tech companies moving into the banking space. However, less attention has been paid to the changes taking place within wholesale and corporate banking, where the digital revolution is also making its presence felt. 

Last year, digitisation was given an extra push by government-enforced lockdowns to stop the spread of the Covid-19 pandemic. This was probably one of the biggest catalysts for bank digitisation since the widespread use of the internet began in the late 1990s. Almost overnight from March 2020, banks were forced to send most of their staff to work remotely, and the majority of their clients were doing the same. 

“So, you’re shifting people on to self-service digital platforms that we actually think are genuinely a better service for them,” says Stuart Riley, global head of technology and operations for the institutional clients group at Citi. He notes that the pandemic saw a revival in clients shifting to Citi’s existing online services. Pre-pandemic clients would query where a payment was in the cycle by telephoning the bank’s call centre; most now do this online. 

Mr Riley explains that before the pandemic some clients could not fully use Citi’s online services because either the company or the jurisdiction in which it was located did not recognise the legality of digital signatures. But that changed as it became impractical to use physical signatures, which resulted in efficiency improvements users are not keen to relinquish.  

“In the past banks lived with certain [operational] weaknesses. With Covid-19 they couldn’t do that anymore,” says Daniela Rothley, a partner at LPA Group’s consulting practice. With staff forced to work remotely, she says banks had to make sure processes were stable, which made them think more about digitisation and automation. 

One area where banks paid more attention was trade finance, which involves a complex trail of documentation, much of which is not standardised. Many banks have started digitising much of that documentation, making workflows more efficient and resilient.  

Nervous supervisors 

Supervisors were also a direct influence on greater digitisation during the pandemic. For example, Germany’s Federal Financial Supervisory Authority, as part of a European Central Bank audit, has been probing banks’ operational resilience more closely than usual – possibly due to bad memories associated with the Wirecard scandal – with fears of missing something important.

“This was really a tough time for a lot of my contacts. The result was to set up additional digital projects,” says Ms Rothley, adding that these projects often involved automating internal processes to reduce the possibility of human error.  

Indeed, it is not as if banks had been complacent about digitisation pre-pandemic. Quite the opposite: the regulatory reforms triggered by the 2007-9 global financial crisis forced banks to implement new IT systems to collect, analyse and report vast amounts of data demanded by regulatory frameworks, such as Dodd-Frank in the US and the EU’s Markets in Financial Instruments Directive II. This has certainly had an impact on wholesale banking, where manual processes were digitised and automated out of sheer necessity.

But digitisation is far from easy for banks. Moving operations into the cloud can create significant efficiencies, but banks face plenty of hurdles in doing so. 

“The complication comes when you’ve got 60-year-old technology that is highly complex or supporting products that have been going for the past 30 years. How do you migrate those into the cloud? And how do you do it in a safe way that actually delivers value? This is not a trivial exercise,” says Dean Jayson, head of UK banking at Accenture.

Citi’s Mr Riley cites the example of clients who discuss data analytics and research with the trading and sales department and now, increasingly, access the information online. They can import that data into their own systems and more easily use it internally. 

Société Générale has shifted 80% of its infrastructure into the cloud. It is now looking to forge a competitive advantage from the data it has to collect to satisfy requirements such as the Basel Committee on Banking Supervision’s 239 data standards and the EU’s General Data Protection Regulation (GDPR). This data can also support artificial intelligence (AI) and machine learning programmes. However, non-GDPR countries, such as China, can go further with AI as they may have fewer restrictions on the access and use of personal data, which may inform lending decisions, for example. 

“At a large-scale industrial level, we need to implement a clear AI and data roadmap, in order to understand where the pools of data are, which are the most important for us,” says Claire Calmejane, chief innovation officer at Société Générale, adding that this data can be combined with external data to drive new insights for the bank and its clients.

“Today we have about 80 different user cases for AI and 250 for data and AI,” she says. She explains that these initiatives range from automating certain aspects of advisory services and pricing through to improving response times to client enquiries. Société Générale uses chatbots for some retail services in the French market, which Ms Calmejane explains involves processing some 15,000 conversations a day and has earned a 98% satisfaction rate by users. 

Automation, digitisation and AI are used extensively by the bank to meet know your customer (KYC) regulatory requirements and to make the process more efficient. Essentially, this involves documents being digitised so software can extract the necessary information, which then becomes part of the workflow. Similar technologies and processes are used for extracting critical information from legal documents. 

Chasing efficiency 

“You can massively increase the efficiency of the business just by immediately extracting information [from documents] and putting it into the system,” says Albert Loo, global deputy head of sales, global markets at Société Générale. 

This involves everything from optical recognition technologies through to natural language processing to compile and analyse structured and unstructured data. AI is used to extract trading ideas from the bank’s research so it can be more quickly forwarded to clients.

Accenture’s Mr Jayson notes that the way documents are created, accessed, viewed and edited within banks has completely changed thanks to digitisation, which has also affected areas such as reporting.  

Paula da Silva, head of transaction services at SEB, explains that the vast amounts of data required to keep the Swedish bank’s house in order and stay compliant requires continued automation. However, she reflects that it also enables banks to innovate by using that data to create direct customer value. Most of the data that has been cleaned and put to work in automated processes for regulatory reporting can be re-used elsewhere. She cites the example of producing annual statements for customers, which was automated several years ago. “That was a pretty big undertaking, as it often goes across several different product areas and geographies, and is really cumbersome to do,” she says. “Now most of that process is automated.” 

Another area that has benefited enormously from automation is payments. Ms da Silva says it used to involve hundreds of people doing internal registration, processing and verification. “That’s all automated now. Customers can get reports on transactions and balances automatically at their convenience and channel of choice,” she says. That means that as well as SEB’s customers getting a more frequent and accurate real-time view of their liquidity and funds availability, the bank is also able to monitor and forecast its positions more efficiently. 

In terms of AI, SEB runs a wide range of projects at different levels of sophistication, from robotised password reset for employees to automated checking of trade finance documentation as well as payment screening and monitoring. In other cases, AI programs are extensively used for pattern recognition, specifically to fight fraud and financial crime. This pattern analysis is also useful to the client so they can detect trends in their own business.

Ms da Silva says SEB is increasingly thinking about how it can use its vast data resources to support clients. For example, if a client wants to start exporting to China, it is very likely that SEB has substantial knowledge about the Chinese market given other clients that have long been exporting there. Enriching that data with data from external sources would allow the bank to better advise customers. 

“There are a lot of use cases where AI is being deployed by banks currently, such as fraud detection, trade surveillance, anti-money laundering/KYC and operational risk management,” says Neha Singh, vice president, innovation and growth at Broadridge, a technology solutions company. She says one area where AI is being deployed in capital markets is the search for optimal liquidity pools as part of pre-trade analytics. “This can help enhance trading decisions,” says Ms Singh. She believes that a competitive gap is already developing between those banks that are enthusiastically adopting AI and those that are lagging behind in adoption. 

Governance keeps pace

“At group management level, we trained all our senior executives to explain how AI works, to understand the different types of machine learning, whether it is supervised, non-supervised or reinforcement learning,” says Ms Calmejane. To ensure these programs perform as expected, she says the bank relies on traditional risk governance approaches. 

If someone within Société Générale wants to launch a new AI product, it has to go through a product committee. The program’s model is assessed, as are the risks and governance issues. Meanwhile, the senior executives involved have to understand the technologies behind the application.

Mr Riley says Citi takes a prudent view over allowing AI to make decisions. “We’ve got quite an extensive internally defined framework as to how we think about machine learning and AI, what category it falls into, and therefore what kind of governance we need around it,” he says. He adds that Citi does not allow AI to make decisions, particularly if it affects customers, and instead the output is taken and is assessed by a qualified professional. “And we do that because, I think rightly, we’re mindful of the ethical or moral implications there might be in the data feeding those models,” he says.

Digitisation is only part of the story of rethinking banking. Currently, AI and machine learning are mostly used for KYC, fraud detection, revealing market abuse and onboarding clients. But looking ahead, bankers see many exciting possibilities for creating increasingly personalised user experiences for their clients and more operational efficiencies supported by the convergence of a range of technologies, such as cloud, AI, blockchain and quantum computing. 

This article first appeared in The Banker’s sister publication Global Risk Regulator.