The Future of Banking on artificial intelligence

Pioneering financial institutions are exploring where AI can make the most difference to their operations – from customer service to fraud detection.

Artificial intelligence (AI) is the latest in a long line of technologies to play a part in the digital transformation of the financial services industry. The potential of this technology is vast: it can cut costs, provide human and systemic efficiencies, boost customer experience, promote loyalty and boost returns.

According to business research firm Gartner, the two key components of AI – machine learning and deep learning – will be adopted as the norm within the next two to five years. There is real impetus and enthusiasm for organisations to adopt these technologies.

Andy Pardoe, principal director of AI at Accenture Digital UK&I, says: “AI can be used across the entire value chain, from first contact with a potential new customer all the way to providing additional services to long-term customers. This is happening across the front, middle and back office functions.”

Essentially, AI is a series of underlying technologies: natural language processing, computer vision, machine learning, deep learning, neural networks and others. These are all brought together within a cloud-based environment that can store and process massive quantities of data and allow for instantaneous AI interactions.

Pardoe says: “AI is a broad term that covers a multitude of techniques, from simple rules-based methods through to natural language processing, that uses deep learning. The main focus with AI at the moment is with a subset of techniques that fall into the machine-learning category. These all fundamentally work by leveraging data to learn from it.”

Isaac Ben Akiva, head of machine learning at Barclays UK, explains: “I think it’s important to distinguish between artificial intelligence, which is about reproducing human cognitive capabilities in machines, and machine learning, which is about teaching machines to identify patterns within data. The two are certainly connected, and machine learning has undoubtedly been one of the most important advancements in the development of AI in recent years.”

Machine learning, therefore, deals with the data which can then be built on by cognitive techniques, which deals with actual interactions with humans using the underlying data to make decisions. It can extend the capabilities of both humans and machines far beyond what each can do on their own.

Didem Dinçer Baser, executive vice-president in charge of digital banking at Garanti Bank, adds: “The increase in our capability to provide and process data and the development of AI creates new opportunities for us. Big data and AI are powerful technologies that will change the nature of banking because we are now moving towards a learning and self-improving structure where all processes can communicate in real-time with each other. It will improve our trust-based relationship and enable our customers to have more meaningful interactions with us.”

The capabilities of AI are not in doubt. But where can it be used for greatest effect? Thus far, process automation and data mining within operational efficiency and risk management have been the easy targets.

Initially, AI capability was seen as an enhancement to data analytics. The idea was that the machine-learning component of AI would provide better processing of middle- and back-office data and reduce human intervention and thus potential for error. The intention was to optimise and automate and it is here that the predictions of AI replacing humans are more prevalent.

Analysis published by Forrester Research in 2015 estimated that, by 2019, robotic automation would change up to 25 per cent of the work associated with all job categories.

Accenture Digital’s Pardoe says: “AI can make the most difference in several ways. First, it has the ability to help automate the mundane while improving efficiency and quality for highly repetitive tasks. This enables humans to be freed from these robotic tasks to focus on higher value, more complex and creative work.”

Examples include JPMorgan, which has successfully used AI in this context to make its trade execution process more effective. Citi has also developed its machine-learning capabilities to support its pricing requests that are sent to traders. This is supervised machine learning.

Avika of Barclays UK says: “With supervised machine learning, you give the computer a set of data and tell it to calculate a specific outcome. With unsupervised machine learning, you’re asking the computer to identify patterns without knowing what the correct answers will look like. For example, you might ask the computer to group your customers into different segments based on their behaviour, allowing you to better target your marketing activity.

Unsupervised machine learning, meanwhile, has been successfully applied to the compliance and risk process. Here, the idea is to identify the needle in the haystack. The technology scans data and documents and takes action based on a set of laws and regulations or parameters.

HSBC is one example of this. It is applying AI to its money-laundering, fraud and terrorist-funding detection process.

Another noteworthy example is Singapore’s OCBC, which partnered with BlackSwan Technologies and Silent Eight in 2017 to improve and speed up its ability to detect and investigate suspicious transactions.

Akiva says: “Machine learning is being used in fraud detection and prevention to spot anomalous behaviour in real time, reducing the number of transactions that need to be escalated to a human for enhanced due diligence and, ultimately, helping keep customers’ money safe.”

This is closely linked to Regtech (regulatory technology), where regulation becomes automated, which the UK’s Financial Conduct Authority is actively encouraging via its Fintech Sector Strategy.

Singapore’s DBS bank offers an example of support for Regtech: the latest round of its Accelerator programme includes CUBE, a UK-based Regtech company that uses AI, machine learning and natural language processing to capture global regulatory data. This is done automatically and on a continuous basis, creating a data map of cross-border regulatory intelligence which can then be applied to a bank’s regulatory process. That lets the bank identify where regulatory touchpoints are in terms of jurisdiction and line of business. Anything new is automatically flagged and managers then know where to act.