7 minute read

Let’s talk about the next wave in AI, Machine Learning & Managed Services

SmartStream’s fully integrated suite of solutions and platform services for middle- and back-office operations are more relevant than ever – proven to deliver uninterrupted services to critical processes in the most testing conditions. Their use has allowed our customers to gain greater control, reduce costs, mitigate risk and accurately comply with regulation.

With AI and machine learning growing in maturity, these technologies are now being embedded in all of our solutions and can be consumed faster than ever either as managed services or in the cloud.

Simply book a meeting to find out why over 70 of the world’s top 100 banks continue to rely on SmartStream.

info@smartstream-stp.com smartstream-stp.com

The use of AI by financial institutions is growing inexorably, allowing a multitude of tasks previously undertaken by armies of back-room staff to be handled both more efficiently and at far greater speed and scale via automation.

That’s not to say, automation always displaces people. More often, it does the heavy lifting in a new kind of ‘job-share’ between machine and human, augmenting intelligence rather than substituting our own with the artificial.

But, as the technology becomes more sophisticated, will it, ultimately, do away with the need for ‘us’? Not according to Ash Booth, head of AI, Markets and Securities Services at global bank HSBC, who argues that human intervention is now needed more than ever to maintain one vital commodity: truth.

Here, Booth, who started his career in academia and holds a Phd in AI applied to finance, shares his thoughts with The Fintech Magazine’s Doug Mackenzie.

THE FINTECH MAGAZINE: How is HSBC using AI to relieve staff of repetitive jobs and add value to services?

ASH BOOTH: Many of the recent developments in AI can lend a hand, not just in terms of saving costs, but more enabling scale that was previously bottlenecked by just the amount of work that people could do – we get tens of thousands of emails, instant messages, faxes, unstructured SWIFT messages, phone calls from clients, constantly.

There’s been much in the press recently about generative AI, but there are also developments in image recognition and language modelling that can hugely help us in processing client queries. We want to be on top of the latest developments to ensure we can scale. But, importantly, too, we are a bank that prides itself in its relationships. We want to be a bank that is there, in person or person-to-person virtually for our customers.

So there’s an interesting balancing act we have to play; we need to automate to scale our business, but we need to be there for our customers. And we’ve been very surgical about how we’ve used automation over the last few years to do that – and now, especially, that we’re seeing some incredible advances in those spaces. That could be through chatbots, but often it’s the stuff that happens in the background.

It’s important for a lot of our client bases that they feel like they are – and often they are – speaking to a real person, but there’s so much we can do around that interaction to make the client’s life easier and also to make the jobs of our operations staff, or our salespeople, or our client services staff, easier – to automate some of the really mundane, lower-value things, so that they can have the higher-value conversations. something from scratch. I think, over the next few years, it’s going to become the norm to have something automatically generated for you, and you will revise that for many reasons – I already mentioned proving correctness and truth worthiness.

Look at the sales and trading space, where there’s a lot of exchange between our traders, our salespeople, and our clients. Increasingly, that is electronic, but I think people are often surprised to hear that a huge chunk of that – and, in many asset classes, the majority of the trading by volume – is still done ‘by voice’. By that we mean over the phone, but also via instant messaging, like Bloomberg, or Symphony, or Reuters. It’s very important, but it’s often very time-consuming, and really quite transactional.

If we rewind a few years, you’d often hear a salesperson, who was dealing with the client, shout over to a trader, who was sitting next to or opposite them, to get a price. But, more recently, regulation has meant, no matter whether the client trades with us or chooses to trade with someone else, we have to log that interaction.

We have to say that client X asked us for a price, we gave a price, and they either traded with us or chose not to. So, that created a big push to electronify that person-to-person interaction.

We turned to natural language processing and got to a place that maintains that client/customer interaction, and person-to-person chat, but with AI systems in the background automatically detecting that the client is asking for a price on something else, creating a ticket in the system, showing that to a salesperson, for them to check, and verify, then push the information back to the client.

So, we’ve been able to maintain that person-to-person interaction, to meet our regulatory obligations, get some nice management information off the back of it, and, importantly, go from a minute to get a price back to a client, down to seconds.

TFM: What does HSBC do to enshrine ethics in its use of AI?

AB: We have a very well-communicated and established set of ethical principles, specifically now around the use of data and AI, and governance structures in place to make sure we align to those.

Each of our businesses has an AI ethics committee and people who want to experiment with, or are thinking about using AI to capture an opportunity, or solve a problem, will bring that idea to the ethics committee – which is made up of senior leaders – walk them through the use case, the data that they’re hoping to use, also the tools and techniques that they’re hoping to employ, and their aimed-for outcomes.

The leaders of that business can think about any potential ethical implications, and are now also completely aware of everything that’s going on, and have some informed discussions about not just how we should do things, but whether we should do them, whether it’s in line with our ethical values.

Once we’ve started on development, my team, in conjunction with a few others around the group, maintain a set of best practices for fair and transparent AI.

TFM: With technologies like ChatGPT now becoming mainstream, what will that do for people’s expectations of chatbots?

AB: HSBC, and many other firms like us, explored, or full-on adopted, chatbots, specifically in the retail space. It helps scale customer services while maintaining that personal touch, when done well. And, at HSBC, we transformed some of what you might call traditional operational staff to more chatbot conversational designers.

This buzz around large language models that are powering ChatGPT, and alternative systems, present a really interesting opportunity, but also a challenge. The output of the ChatGPT and similar models is incredibly impressive, on the face of it, but there are certainly concerns around how we guarantee truthfulness.

It’s easy to forget that these models aim to generate plausible text, in their current form. So, for those of us who care very much about truth and correctness, we have to think very carefully about how we incorporate those models into our existing systems. There’s certainly a lot of potential but we need to do a lot of work to understand how we get that power of fluid conversation, of a really pleasant interaction, but be able to guarantee correctness?

Also, some of the big players in this space have been quite open about their plans to integrate with a lot of the common productivity tools that we use. I think that will do wonders for productivity, and it will set a general trend for how we think about integrating AI into some of our strategic systems.

Often, we naively have this idea that AI will replace a thing, or replace a process, or will be quite in your face – you interact with an AI system. I think it will be much more in the background with a lot of mundane analysis done for us to guide us on our way so that we can focus on those high-value conversations, or trades, or complex risk management processes.

TFM: HSBC has become a very big technology player. Will it look at developing more AI services in-house, or partner with tech giants to provide best-in-class services?

TFM: Speaking as someone who’s come from an academic background, where do you we see future AI innovation taking us?

AB: I don’t like to be in the business of predictions, but I have a lot of observations that point in some inevitable directions. With reference to large language models, even if I just look at language modelling – and I don’t speak just for financial services here – I think we’re on the cusp of a transition, in terms of how we work, from what I could call the era of generation to the era of revision.

Think about how you might use, and how many people are already using, things like ChatGPT. Whether you’re a programmer, you work in marketing, or you’re writing research reports, in many jobs you start with a blank page and you generate

AB: I think a lot of the productivity gains we see will come from many of the various technology partners that we work with, providing additional services to a lot of the tools we use. We encourage that, we want to see it. We’re going to have to work really closely with those firms, to make sure that they are aligned with our ethical principles, how we want to use AI, and that’s a challenge, too.

On the other hand, when it comes to our kind of core value proposition, when it comes to how we manage risk, how we price things, how we measure the probability of certain risks, that’s an area where I see us leading, and developing a lot of things in-house.

The wonderful thing about the AI academic community, and even some of the tech giants, is that there’s a culture of openness around knowledge. We see a lot of people across the business writing academic papers, speaking at conferences, it’s a very open community and that’s good because, fundamentally, banking is the business of managing risk, and although, of course, we want a competitive edge, if everyone is better at managing risk, I think that’s good for us all.

This article is from: