Contextual banking is here, and AI is enabling it

Artificial intelligence is an inescapable part of our new reality, reaching even into places we might not realise. It’s a product of globalisation, but also an enabler of it. It drives emerging technologies while helping refine and expand the capabilities of existing ones, and it’s helping humanity address social challenges. The financial services sector is no exception.

What do we mean when we talk about AI? At its core, AI seeks to simulate human intelligence and human traits and skills — like the ability to learn and problem solve. Specific applications of AI include natural language processing (NLP), speech recognition and machine vision. Consider, for instance, smart speakers that can respond to natural human speech, not just a single set of pre-defined commands, and services like Google Photos, which can recognise objects in images it hasn’t previously seen.

AI also encompasses subcategories like machine learning, where a computer is trained to analyse, interpret and identify patterns in large datasets, and then to make decisions from that with little to no human intervention. Deep learning seeks to mimic the way the human brain works to learn or extrapolate from different data models, enabling connections to be made between seemingly disparate datasets.

Spending on AI is enormous, and growing fast

According to SPD Group, banking, financial services and insurers’ combined spending on AI will climb to US$12-billion by the end of 2021. But that’s just the start. McKinsey expects the industry to eventually grow to a trillion-dollar one in coming years, with two-thirds of that value coming from traditional AI and analytics being harnessed by marketing, sales, risk and HR, and a further third coming from advanced AI functions, mostly in the risk space.

But how will financial institutions meet the challenges of integrating AI into core parts of their businesses? By embracing disruption rather than innovation. Innovation tends to mean doing what you’ve always done but doing it a little better than before. What’s needed instead is to create systems and processes that make old solutions obsolete.

When people talk about “disruption”, they often actually mean innovation that derives from first-principles thinking. That is, the sort of reinventive innovation that stems from going back to the drawing board. SpaceX’s Falcon Heavy rocket is a perfect example: By dispensing with the notion that rockets had to be disposable, Elon Musk’s company executes launches to space for a tenth of what it used to cost Nasa.

The danger of using new solutions to do what you’ve always done is that, because the inputs are the same, your outputs will be similarly homogenous. That’s why more than 80% of companies fail at digital transformation: They see the shift to digital as evolutionary, when in fact it needs to be revolutionary.

The AI-powered bank of the future

Digital onboarding isn’t just a nice-to-have for customer acquisition, it’s now something customers expect. AI and ML can add huge value to the know-your-customer process, cross-referencing documents, verifying identity and evaluating risk.

As Uber proved, it can also help with acquiring otherwise unreachable customers. The ride-sharing business found drivers in some markets were struggling with onboarding because they were unbanked, so it partnered with BBVA bank to create accounts for new drivers and used AI to link the newly created banking services with their Uber accounts.

It’s estimated 330 million new accounts will be opened via online banking channels exclusively by 2025, almost double the 184 million opened that way in 2020. Using AI in conjunction with biometric verification or other self-service tools removes friction and improves efficiency, and could save financial institutions almost half a billion dollars. By 2030, banks will have 2.5 billion new customers, 94% of whom will never have been to a branch.

But verifying ID isn’t enough. Customers want businesses to know not only who they are, but how they behave, and to preempt their desires. And they want all of this without any security compromises.

Amazon Go stores, for instance, use cameras and biometrics to see what customers take from shelves and then bill them automatically. Not only does that offer a more seamless shopping experience, but it also allows the retailer to create an increasingly detailed overview of its customers’ habits, enabling more targeted promotions down the line. Customers, meanwhile, don’t have to worry about cash, payment cards, passwords or Pins.

ML improves risk detection and monitoring

Compliance and risk management are inextricably linked, and ML can help with both. Compliance costs the banking sector US$270-billion annually and accounts for 10% of operational spending. Fraudulent transactions for banks, e-commerce businesses and airlines are similarly expensive, costing an estimated $200-billion between 2020 and 2024.

ML can help tackle fraud and identify theft and other malfeasance by spotting behavioural discrepancies or unusual patterns gleaned from myriad data points. Plus, ML systems’ predictive abilities improve because what is learnt from previous instances is applied to all future ones.

At the same time, AI can help automate or augment human risk-management processes. For example, First National Bank’s risk segment developed an AI system called “Manila” that allows the bank to meet regulatory requirements and make forensic due diligence decisions faster, more accurately and more efficiently. On average, the use of AI frees up 70% of analysts’ time, and generating a forensic synopsis ready for a human analyst to review that previously took hours can now be completed in as little as eight seconds.

JPMorgan Chase, meanwhile, uses its own ML-powered system called “Coin” (contract intelligence) to automate mundane and previously time-consuming tasks, and to process credit agreements in seconds rather than minutes, while also correlating the number of loan-service mistakes. The company intends to add ML to other areas, like custody agreements and credit default swaps, because of Coin’s success.

Behavioural risk management

It’s not only threats that AI and ML can help mitigate; they can also help with behavioural risk management. Reinforcement learning (RL) is a machine-learning approach that points to the right action to take to maximise the reward in a particular situation. For instance, a customer’s transactional or investment behaviour may show they’re focused on saving for their children’s education, housing, travel or health needs. A financial institution that recognises this can shape its products or services such that the customer is incentivised to adjust their behaviour to match their desired outcomes.

The same technique can be applied for credit risk, model risk, climate risk and even customer desirability. Instead of risk compliance questionnaires at onboarding, behavioural frameworks can help manage risks on an ongoing basis, while also proactively capitalising on customers’ changing desires or expectations.

Digital engagement via mixed channels

AI, multilingual chatbots and digital voice assistants are changing the customer service landscape, improving efficiencies, and reducing overheads. They enable 24/7 customer support and allow institutions to operate in markets where multiple languages are spoken. In the banking sector, up to 90% of customer interactions can be automated using chatbots, improving productivity while cutting costs significantly and freeing up support staff to work on more complex problems.

At the same time, robo-advisory services are gaining consumer acceptance and helping automate investment decisions aligned to customer needs or aspirations. They can even rebalance portfolios over time, optimise for tax obligations, buy fractional shares or manage tax-loss harvesting. One of the most successful businesses doing all the above is Betterment, which handles retirement planning and other investment services for over 600 000 clients in the US.

The publication BI Intelligence estimated robo-advisors would be managing over $8-trillion worth of assets globally by 2020. Like chatbots, they’re available whenever customers want them, they free up human advisors to work with clients with more niche or complex requirements, and — perhaps most importantly for customer acquisition and retention — millennial customers tend to trust them.

The rise of the robo-traders

Similarly, robo-traders are playing an increasing role for investment banks. UBS has introduced two automation services for clients. The first deals with post-trade allocation requests by automatically scanning customer emails about preferred fund allocations and then executing the transfers. The second uses ML for creating trading volatility strategies. In the case of the former, what used to take a human 45 minutes now takes a robo-trader two minutes, freeing up bankers to call clients and perform other human-centric tasks.

AI and ML are also expanding their capabilities thanks to advances in the computing systems that underpin them. Quantum computers — which harness the unusual characteristics of quantum mechanics to perform computations — can process massive calculations currently only possible with a handful of supercomputers, and far more rapidly.

With ever more computers and devices connected to one another, opportunities and cybersecurity risks grow in tandem. Quantum computing will enhance banks’ ability to provide the sort of encryption needed to defend against ever smarter hackers and malware threats while also making it possible to wrestle with ever-larger datasets and extract insights and value from them.

Contextual banking in the AI era

The core value proposition for banking for generations has been the interactions that take place between savers, banks and borrowers. Traditionally, banking growth has relied on iterating upon the ways these three interactions with one another. But globalisation, changing technology and social challenges require more than iterating on old value propositions or current technology to remain relevant, responsive and profitable.

To meet the needs of future consumers, compete in a growing marketplace and stay ahead of evolving risks, banking will have to embed the legacy value proposition and channels within emerging technologies like AI and ML to spearhead contextual banking. Customers won’t accept anything less.

• Dr Mark Nasila is chief analytics officer in FNB’s chief risk office