Banking on an AI future
Illustration: David Senior
Share on LinkedIn
Share

Banking on an AI future

Jim Mortleman — August 2017
Artificial intelligence looks set to transform the financial services industry. Technology leaders at three major banks — UBS, ING and Lloyds — reveal how they are approaching the transition.

For the fiercely competitive financial services sector, artificial intelligence-based (AI) automation holds the tantalizing prospect of not just unprecedented productivity gains and customer service improvements but also business model reinvention. But where some see opportunity, others see threat. There is growing public concern that the rise of AI threatens countless jobs and undermines individual workplace autonomy like no other technology before it.

Whether they’re looking at AI-powered chatbots to improve service and operations, intelligent algorithms to bolster security, trading, business analysis, or the automation of manual processes, banks know they must introduce these systems with appropriate levels of sensitivity and caution. And while all three IT leaders we feature below agree on many points — in particular, the need to reassure employees that AI will help them do their jobs better rather than replace them — they differ significantly in terms of the pace, extent and nature of AI’s adoption.

UBS: Act now but don’t rush  

UBS — Switzerland’s number one bank — is developing its interest in AI technologies through its established innovation channel, starting at small-scale and cautiously considering wider deployment only when a particular aspect of the technology is deemed sufficiently mature.

That AI program was kicked off in early 2016 after the bank had observed a surge of venture and financial services industry funding of new applications of AI, says Annika Schröder, director of group innovation. And it has cast the net wide for insight, she says, “sourcing ideas from employees all over the world, holding hackathons and focus days, and talking to start-ups and strategic vendors.”

Each of the resulting shortlisted application proposals was presented to the innovation board, which is made up of IT and business stakeholders from all of the company’s core divisions. “Like entrepreneurs seeking backing, people pitch for funding with the understanding that no project should take more than three months or need a great deal of money,” explains Schröder. Judging from the take-up rate, most of those ideas were pretty strong.

AI-driven applications and services now account for about a third of the projects in UBS’s total innovation program, which is also looking at the application of other disruptive technologies such as blockchain and IoT. Potential use cases for AI run across numerous business areas, including customer services, operations, risk and compliance, investment and trading, and cyber-security.

Results so far have been mixed. The company sees potential for chatbots in customer services, for example, but says they’re not yet ready for widespread deployment. “We’ve run trials internally where accuracy of understanding has only been 60% to 85%, which for us is not yet good enough to let them loose on customers,” says Schröder.

Intelligent agents for investment research are showing more promise, but again aren’t ready for prime time. AI agents can spot behavioral biases among portfolio managers, for example, but such applications are still just experiments, she says.

On a project-by-project basis, the technology is nonetheless percolating into live environments. For example, the company is using digital reasoning technology on many millions of messages a day to identify fraud attempts. And in operations, the first project to move from the pipeline into production was a machine-learning tool that can read IT support tickets and find the best way to solve them.

Schröder stresses the technology is not a threat to employee activities. “There’s a lot of irrational fear about AI. People are still key to our business — their creativity, their skills in managing projects, designing, planning and validating systems. I think the relationship between robots and humans will be a collaborative one. So we’re having sessions with the workforce to explain the technology and hopefully alleviate any fears they may have.”

And her best piece of advice to other companies that want to explore the technology? “Do it now. There’s no big risk in trialling things. The technology still needs a lot of development, but it’s reaching critical mass right now and you can’t ignore it.”


ING: AI for the people

The approach to AI at Netherlands-headquartered global bank ING is notably people-centric: it focuses on projects that can not only get the buy-in of senior management but also win the backing of employees and customers.

“Any AI project has to show how it is going to move the business forward,” says Vinoth Raman, global head of robotics and AI. “It has to be positioned as a way of improving end-to-end processes and creating compelling new products and services. If you can find use cases which make a big transformation and increase visibility of the technology’s benefits, management is likely to have an appetite for it.”

One example at ING involves the replacement of a traditional rules-based anomaly detection system with one powered by machine-learning algorithms. Previous testing has shown this will improve performance significantly — much more than the 5%-10% typically attained in a technology upgrade. And that gets managers interested, says Raman.

Top-down support is vital because organizational changes are needed if such technologies are to be deployed to best effect. At the same time, bottom-up backing from employees cannot be overlooked. To ensure that, ING has built a ‘community of excellence’ — a federated model through which the bank’s various divisions across the world can contribute and work on ideas for innovative projects, so everyone feels they can help shape the bank’s use of these technologies.

You also need to win customers’ hearts and minds, says Raman, especially if you want to make use of their valuable data. “Our focus is to attract more primary customers so that we have more transaction data. That data is key for developing compelling AI products and services,” he says.

And rather than simply, say, adding chatbots on its own banking website, ING believes it has to go where customers are — such as popular social platforms like Facebook or WhatsApp — with bots being developed that can interact with customers on that new banking turf. “In that way, customers don’t have to go to a different channel to use services or products,” says Raman.

Whatever AI-based systems make it into production, it is vitally important they work as people want and expect them to, says Raman — if you are to keep people on board. If you have a global clientele, as ING has, you need to ensure your AI systems can understand any language spoken by your customers, he says, and that adds to development complexity.

Whether the stakeholder is a manager, an employee or a customer, focusing on people will be the key to allowing ING to accelerate its progress with AI and differentiate itself from competitors, says Raman: “AI won't replace people, it will enhance their value. But you need them to work together if you want to add the most value.”


Lloyds: AI at a gallop

Sitting at the hub of European financial services, UK banks like to think of themselves as leaders in innovation, and Lloyds Banking Group is certainly a good deal further along the AI road than many of its counterparts. The group, which includes the well-known financial brands of Lloyds Bank, Halifax, Bank of Scotland and Scottish Widows is charging ahead with plans to put machine intelligence at the center of many parts of its business.

“We believe this will fundamentally change how work is done and decisions are taken in the bank, as well as resulting in an experience for employees and customers that’s much more personalized, relevant and contextual,” says Marc Lien, director of digital development and applied sciences at the banking group.

Lloyds is deploying complex cognitive technologies, machine learning algorithms and intelligent process automation across its business operations. “This is no longer about doing a few experiments in the lab. We’re now scaling this up to enterprise deployments,” he says.

For example, the company is improving the functionality of its customer webchat support by piloting intelligent chatbots. “These smart assistants will resolve customers’ queries or, if they’re unable to help, pass them on to a human operator,” says Lien.

But in order to improve their ability to understand customers’ intentions and resolve their issues, chatbots are initially being deployed internally to help the bank’s employees answer customers’ questions faster and more effectively. The bots get smarter as they learn how to solve the various issues presented to them, reports Lien. “We have tens of thousands of employees dealing with customers — from client managers to telephone operators. The chatbots draw on the bank’s entire corpus of information and bring back anything relevant to employees as they need it. And as we improve the interface, we’ll start rolling out the technology to customers,” Lien says.

The bank makes clear that its goal for AI is not to replace human employees, but to help them do their jobs better. “For us, it’s about augmentation rather than pure automation pairing brilliant people in our business with increasingly smart technology to deliver great things.”

This line of thinking continues into the bank’s approach to fraud detection — already a classic focus for the application for AI. Instead of simply focusing on algorithms, Lloyds is looking to use AI to embed intelligence into the customer or employee workflow.

Lien also highlights opportunities for productivity enhancement. “We’ve built our architecture so we can plug in lots of cognitive and machine-learning components, largely drawing on open source tools and projects.” One early example his team has built for its own use is a tool that speeds up software development.

“We ask customers to give us feedback on beta releases [of online products], and that resulting feedback requires a lot of deep analysis. Previously we had a team reading through it all to understand what was and wasn’t working as required. Now the tool intelligently analyzes the feedback in real time and gives us a good sense of customer sentiment at the point of deployment,” says Lien. That means product owners don’t have to wait days or weeks to make changes and move on to the next iteration of product release, he adds.

Lloyds has learnt plenty of lessons on its journey but perhaps the most important is to adopt the right mindset. “You have to understand AI isn’t something that can be tacked onto the side of the organization or left with a business innovation team to do on their own. It needs to be a core part of any transformation. You have to move at scale, right across the enterprise,” says Lien.

• Annika Schröder, Vinoth Raman and Marc Lien were speaking at TechXLR8 in London.
First published August 2017
Share on LinkedIn
Share
I-CIO on Linkedin
    IMPORTANT:
    This website uses cookies to improve functionality. Continue to use the site as normal if you're happy with this or click here to find out more about cookies.