Aditya Challa, IMImobile: choose the right technology and use AI wisely!

Aditya Challa, IMImobile: choose the right technology and use AI wisely!

Aditya Challa of IMImobile takes a look at some current and future applications of artificial intelligence (AI) for customer experience in the financial and banking sector, as well as the challenges that still need to be overcome. Beyond the current race to build chatbots, he predicts that voice-driven AI services will ultimately become the main interface between banks and their customers.

Robot banking

When four out of five bankers agree on something it’s worth taking notice. According to Accenture’s Banking Technology Vision 2017 report, that’s how many think AI will revolutionise how they interact with their clients and collect information.

The banking industry has been using automation and AI for years, of course. Anyone who has ever been turned down for a loan or credit card by a computer algorithm will testify to that.

Automation is being driven partly by technological advances, partly by the need to deliver more with less, and partly by customers, who want to use their smartphones and other digital devices to solve their problems in the fastest and most efficient ways. If interacting with an AI-enabled system is faster, the smart money says that customers will embrace it even at the loss of human contact.

Despite the rise of machines, real clarity on how AI-powered services and interactions will be delivered, and the value they will add, is hard to find. Many analysts cite concerns about the “known unknown” and “unknown unknown” problems they could create, particularly around security, consumer take-up, and what the preferred user interface will be.

Virtual assistants and chatbots

The “2017 Retail Banking Trends and Predictions” report is ebullient about the use of AI in the finance sector. The top two trends identified by over 900 global organisations were removing friction from the customer journey, and the use of big data, AI, analytics and cognitive computing.

Banks and financial services companies are already combining those two trends – along with the third trend of improving integrated multichannel delivery – by deploying AI-enabled chatbots and virtual assistants to help both in the back office and in customer-facing situations.

Imagine looking at an item in a shop, and your augmented reality (AR) glasses display a credit offer for it. You can apply for the credit just by talking to your personal virtual assistant (VA), which already knows all your financial information. This is the kind of future we are talking about in under 10 years.

Organising one’s personal finances is complicated, and it is only becoming more so. If VAs do take off – and many of the world’s largest companies are trying to ensure they do – then getting them to help out with budgeting, tracking spending, applying for loans, sorting tax returns, and other finance-related tasks is going to be one of their most popular uses.

Who’s already doing what?

  • AI agents with names like Amelia (from IPSoft), Olivia (HSBC), and Collette (Accenture’s virtual mortgage adviser) are already interacting with customers, enabling them to perform a wide range of tasks digitally without human intervention.
  • IMImobile is developing an Amazon Alexa “skill” as a proof of concept for one of the UK’s high street banks which will allow customers to access account information (such as account balance) and perform certain actions (such as payments) using voice commands. Banks are increasingly rolling out chatbots, too, on platforms like Facebook’s ubiquitous Messenger to start offering automated services.
  • A new app called Finie aimed at banks will allow their customers to query their accounts using natural language questions such as “Do I have enough money to go out to dinner tonight?” or “How much did I spend on groceries?”

Personalising the impersonal

The Accenture report goes so far as to say that AI will shape the primary way banks interact with their customers within three years. That might be a little optimistic, but it’s clear that with the continued advancements in machine learning, the capacity of AI-powered technologies to understand complex problems, gather relevant information, decide outcomes, and intelligently interact with customers is growing rapidly.

Rather than distancing the financial services provider from the customer, the use of AI should, in theory, bring the two closer together than ever before. Automated interactions, whether using chat or voice, generate data – which is the fuel of digital business. Applying big data analytics – a form of AI itself – to this data can deliver deeper and more accurate insights than ever before about an individual customer’s needs, allowing for more personal automated interactions.

For example, personal finance apps already exist that customise themselves to their user’s behavior in order to help them balance their budgets or give advice on stock portfolios. Examples include Mint, Venmo, Qapital and Stash Invest. It’s no great leap to use data-driven technologies to genuinely personalise not just live agent and automated customer interactions, but also service levels and even products.

Developing these kinds of close relationships with customers may also be necessary to fight off competition from new entrants. A new report, “Tomorrow’s AI-Enabled Banking”, from IPSoft found that 73% of millennials would rather trust their finances to tech companies like Google, Amazon or PayPal than to their own bank. When the new Payment Services Directive (PSD2) comes into effect in Europe in 2018, disintermediation of banks becomes a possibility as nimbler start-ups or tech companies offering more innovative customer experiences place themselves between banks and their customers. Only by “disrupting” themselves may traditional banks hold on to their customer relationships.

The simplest UI of all

Banks are already betting heavily on so-called conversational finance by deploying chatbots either in their proprietary apps or on platforms such as Facebook Messenger. With the growing popularity of intelligent virtual assistants, and with the likes of Amazon’s Alexa and Google Assistant becoming embedded on mobile devices, in cars and in homes, voice — not a screen — will likely become an alternative interface for accessing and managing a bank’s AI-powered services.

Customers are already used to interacting with banks through machines, be it an ATM, online banking website, or smartphone app. Advances in AI lead us back full circle to “voice” being the main user interface rather than a screen. Talking, even if it is to a robot, is much more natural to human customers than typing, tapping or clicking.

Not only will banks be able to quite literally develop their own “tone of voice”, but banks will be able to use AI to gather customer data and give highly personalised advice on products such as loans and mortgages, reducing costly human interaction while delivering improved customer experience.

Challenges in an AI age 

Despite the potential, there are a number challenges that established players need to overcome. These include security, privacy, accuracy, fallout from errors, threats from fintech start-ups, and the danger of losing the human relationships we still think are needed to cement a relationship between brand and customer.

  1. Security and privacy

While security will always be an issue with any complex, online system that has to use the public Internet, voice-based AI systems do have certain advantages. Speech-driven voice biometrics – that helps to recognise and verify customers by the pitch, tone and timbre of their voice (and not just via a pre-set phrase or passcode) – will be key to protecting accounts. Banks will also need to consider challenges such as avoiding a customer’s intimate financial details from being overheard as they use voice services. This is where a mix of voice and screen could work well together, and indeed Amazon launched a touchscreen for Alexa, at the Echo Show.

  1. Empathy

While it’s always possible to personalise customer relationships using massive amounts of data, any upside can be cancelled out if individual customer interactions are felt to be robotic. While a virtual assistant that tries too hard to be a human has been shown to be off-putting, some amount of digital empathy needs to be on display. While we like to think our financial decisions are based on cold figures and logic, in truth all human decisions – particularly purchasing ones – are rooted in emotions.

IMimobile’s Alexa skill for one of the high-street banks uses Machine Learning and Natural Language Processing, which means it can maintain context and add personality. Humanisation elements include custom-built dictionaries, which can accurately identify a specific product nomenclature for handling customer conversations, and sentiment analysis.

  1. Accuracy

There is also the simple, yet undeniable fact, that for all its advances and achievements to date, AI based on machine learning, neural networks, and natural language processing is just not quite up to the mark yet. As with self-driving cars, even when digital assistants can handle 90% of cases they encounter, getting them to be 100% perfect – or even as good as a human on average – is where a lot of work still remains.

Most sectors of the financial services industry are regulated, so it is important to eliminate errors and misunderstandings. The best approach here is a deterministic one (where we know what we don’t know), which is more suitable for managing risk than a probabilistic one. A prominent concern about the use of AI, perhaps especially with virtual assistants, is the risk of “type-1 errors” – where a misunderstanding of a voice command by a bot can result in a catastrophic mistake. A rule-based or deterministic approach mitigates the risk of type-1 errors entirely, while type-2 errors can be passed over with context to human agents.

  1. New competition

Advances in technology always open the door for new competition. In the financial services industry this is compounded by upcoming and fairly sweeping regulatory changes like the Payment Services Directive (PSD2) coming in 2018. Once banks are forced to open up their APIs, apps and services provided by other companies – like fintech start-ups – will have a direct line into their users’ bank accounts. This tears down some longstanding barriers to competition and to an extent levels the playing field between old and new players.

This is of course a tremendous boon for end customers, who will have a much greater freedom of choice. As they now do with energy providers, they will be able to shop around for the best value for all sorts of services. Banks may soon find that they no longer own the relationships with their own customers, having been disintermediated in the same way as many utilities companies were.

Innovating using technology such as AI is an obvious answer to the moves that fintech players will be making. However, banks do still have one huge advantage, and that’s data. Leveraging the vast amounts of data they have on customer behaviour and preferences will enable them to protect those relationships by hyper-personalising products and services.

  1. Platform dependency

Customers might move from Twitter to Skype to Messenger to Whatsapp whenever they want, but there are much greater restraints on enterprise IT systems which can cause lags if a company wants to keep up.

Right now, for example, Messenger would appear to be the platform on which to build a chatbot, as Facebook has put together some impressive and sophisticated tools. For a virtual assistant with a voice interface, Amazon Alexa is the current clear choice. But for how long?

Building an AI solution that is based on multiple toolkits – such as MS Azure, NLTK, Stanford, Spacy, and Duckling, with fallback APIs to ensure service continuity – and is thus capable of integrating with multiple consumer-facing services reduces or removes these limitations. As a result, corporations will not be tied to the specific roadmaps of the most popular consumer-facing platforms.

Shape the future

For established financial services players such as banks, there are as many threats arising from technologies like AI as there are opportunities. The impact, when we look back in 10 years, will have been totally transformational, not just on products and services but also on company cultures becoming entirely customer-centric.

Fortunately AI, used wisely, gives banks the ability to personalise like never before, even while reducing costly human interactions – but choosing the right technology is the key to success.