EXCLUSIVE: ‘A new chapter for AI’ – Seán Jevens, AIB and Pierre-Louis Durel, Yseop ‘The Fintech Magazine’

Understanding a story and telling one are very different challenges for a robot. Seán Jevens, Chief Digital Officer at Irish bank AIB, and Pierre-Louis Durel, Head of Customer Success for AI specialist Yseop, compare notes.

The Irish government has just given itself the mission of making the country a global leader in the use of artificial intelligence (AI). Its eight-strand policy, announced in July 2021, has a particular emphasis on AI education, promoting the adoption of AI tools by Irish enterprise, as well as building a strong ecosystem to benefit the economy and wider society with a ‘people-centred, ethical approach to AI developments, adoption and use’.

The strategy builds on the fact that the country’s capital, Dublin, hosts one of the EU’s 30 Digital Innovation Hubs, and that Ireland has long been an important centre for the European operations of AI trailblazing technology giants Google, Apple and Facebook. They, in turn, have helped to create a thriving network of fintech firms, mainly sited in Dublin’s docklands, which is now inevitably known as Silicon Docks. That ecosystem has also fed the country’s major banks, which have collectively invested more than €3billion in their digital services in the last five years.

Irish firms have made up a third of the 100 fintech partnerships forged in the process, according to figures compiled by the Banking and Payments Federation Ireland (BPFI). An example of one of those deals involved Ireland’s largest retail bank, AIB, and the Dublin-based data analytics company Boxever, whose technology the bank leveraged in order to better engage with its customers. AIB has so far used AI in biometric onboarding, to interrogate datasets to find cohorts of customers who might be interested in its services, to retain existing customers, and machine learning in payment tracking and reconciliation.

AIB’s chief digital officer, Seán Jevens, says it has been thoughtful about where it applies the technology. “As with all great technologies, there is always the risk of tech hijack where something has no real, productive use,” he says. AI on its own is of limited value, he points out, unless you have a way of interpreting and delivering the insights it reveals when they are needed – such as in customer relationship management (CRM).

“A lot of the failure I’ve seen around AI and CRM wasn’t because the insights weren’t good and clever, it was due to companies’ inability to deliver those insights to the right people, at the right time,” says Jevens. “That comes down to integrating machine learning into delivery mechanisms. But at AIB we are able to spot propensity triggers and now we’re getting to a point where we’re able to deliver them in a timely way, too, to customers and staff, so that they can act on them.“

While, for banks like AIB, the big focus up until now has been using AI to translate human experience into code, equally important for their back-office processes is translating code into something intelligible for humans. It’s another area in which AI can potentially help to reduce the significant costs of meeting regulatory and reporting requirements. To quantify that, French AI software specialist Yseop, says it has reduced analysts’ time spent writing and updating reports from 48 per cent of their working day to just nine per cent, by applying its pioneering natural language generation (NLG) tech to translating structured financial data into a clearly-written narrative over its Augmented Financial Analyst platform. Pierre-Louis Durel, Yseop’s head of customer success, says it’s suitable for profit and loss analysis, risk and compliance, as well as fund and portfolio reports. And he claims it offers a potentially big return on investment.

“There are a lot of controllers spending a lot of time just consolidating data to do a first analysis. It’s a big advantage to be able to free up this time for them to exploit the analysis rather than just write it down,” he explains. “We have more than 50,000 users live in Europe, the Middle East and Africa (EMEA) and the US – it’s our daily obsession to make data real for banks, insurers and others; to make AI accessible to all.”

Hot on the heels of the release of Augmented Financial Analyst, Yseop also developed an application called ALIX, which quickly assesses if a report is suitable for interpretation by the platform – a kind of AI triaging. The Analyst platform and ALIX, as well as the development of a no-code studio so that it can embed some pre-packaged analysis and text to ease report creation, are just the latest in a long line of investments made by Yseop since it was founded in 2007 by a mathematician and a linguist. It’s much harder for a robot to write a narrative from code than it is for code to be written from narrative. It has to take into account all the randomness of human nature in the interpretation.

“But at Yseop, we have the option to combine machine learning with deterministic AI, to be able to generate error-proof text, and, at the same time, adapt the text to user preferences,” explains Durel. “I think we’ve found the right balance between deterministic AI and probabilistic AI to have the best of the two worlds.”

One among several of the large European banks with which Yseop is working on intelligent automation roadmaps recently approached it with a request to make its reports more readable.

“It was a manual internal report, with a lot of tedious graphs, and it was pretty complex to understand,” says Durel. “The bank decided to use our technology to revamp it with more dynamic text and a focus on the important insights. As a result, this internal report, because it’s so much easier to read, is widely shared today and we are extending the parameter for analysis so that the bank can have more and more people using it on a monthly basis.”

Back at AIB, despite the significant investment it has so far made in its digital services, Jevens is candid that it still has a long way to go in its AI journey. Its chatbot is, for example, still taking ‘baby steps’, he says. But the bank is hesitant for a reason.

“There are lots of bad examples [of robotic chat services]. Customers have been burned by that, provoking comments like ‘I can’t talk to anyone. They’re putting this ridiculous chatbot in front of me that can’t understand anything I say’. So how we stitch that natural language into our core API service is, for me, the real challenge.”

Nevertheless, as AIB shifts towards having more of its products and services available as APIs, Jevens sees huge potential in using natural language processing to make its core services available within key Big Tech apps such as Facebook’s WhatsApp and Messenger, Google’s Alexa and Apple’s Siri. That’s especially so as customer demand for point-of-sale credit increases and the appetite for having multiple apps stored on devices like smartphones almost certainly wanes.

“We’ve already adapted for that, and will continue to do so,” Jevens says. Durel also sees AI propelling banks to ever-higher levels of efficiency and customer experience.

“It’s very clear that our AI has enabled banks to cut costs, but it’s not the only benefit,” he says. “Our customers are insisting a lot on operational excellence – we talk even more about operational excellence than cost reduction. It enables them to standardise and speed up their processes, standardise the documents they are producing to avoid any error, any risk. It’s clearly a big benefit.

“The second important thing for them is the internal satisfaction of analysts. AI is clearly a must-have for providing more capability on a daily basis. I think it could also be a way to retain talent in a bank or other organisation, too.” Jevens agrees that the future adoption of AI in banking will impact on many levels – not just cost reduction.

“I think AI will help us standardise and improve our data and service quality across the board,” he says. “A lot of our digital investment so far has been in the priority use cases, the high-volume customers. Where we really need to stretch digital  now is the more complex cases – the joint customers, the multi-party SMEs – and when dealing with fraud cases, to get to customers more quickly whenever we suspect some activity on their account so that we’re not relying on an agent who’s spotted something, to pick up the phone and ring them. We’ll be much more interactive, in terms of how we do that.

“The other place we see it really helping customers is preventing them falling into debt, by being more proactive as a bank in spotting early signals of vulnerability.

“Australian banks have done some good work in this space, suggesting the customer, for instance, uses spare funds to pay down their credit card. It’s maybe counter-intuitive for a bank, but, in the long term, it’s the right thing to do for everybody. “In Canada, banks have done work in the opposite space, in helping people save, again using AI to identify signals and saying ‘well, actually Pierre’s got spare cash at the end of the month, let’s help him save that’ in an intuitive and natural way. I think that’s where AI will come more to the fore, as we become more comfortable with it.”

And, as it gets more comfortable with us – as adapt at telling as it is understanding human stories.

This article was published in The Fintech Magazine #21, Page 95-96