Artificial intelligence (AI) systems are being regularly used by 18% of organisations, according to consumer information management firm Callcredit Information Group in its Fraud & Risk report.

With take-up of AI on the rise, PricewaterhouseCoopers (PwC) recently predicted that global gross domestic product (GDP) will be 14% higher in 2030 because of artificial intelligence.

This is the equivalent of an additional $15.7trn, or more than the current output of China and India combined, according to PwC research.

Data analytics AI technology is in the early stages of deployment but is already set to transform financial markets and could act as a key differentiator in performance, Matthew Hodgson, CEO of Mosaic Smart Data, tells GTNews.
Two of the key areas within financial services that are making huge advances in AI are fraud prevention and lending and leasing. Over the next three years, 24% of organisations are planning to introduce artificial intelligence for fraud prevention, Callcredit’s report found.

“The trust, however, is not there yet across all segments of asset finance”

AI is also being used to significantly increase efficiencies around new business processes in lending and leasing. In the past 10 years, there has been a major step forward, says Bertrand Cocagne, head of product at Linedata Lending & Leasing.

“In small ticket and auto finance, some lenders are using automated credit scoring to underwrite almost a 100% of deals,” he reports.

“Automatic credit scoring and processing is arguably the most advanced use of AI in asset finance. In theory, masses of back-office data built up over decades around the customer, their assets and payment history should inform the decision on whether to accept or refuse credit.

“The trust, however, is not there yet across all segments of asset finance, and in the more complex and bigger ticket credit cases AI is no more than a guide helping humans validate decisions,” he adds.

Ultimately, trust will follow from a positive experience, says Philipp Shoenbucher, co-founder and chief data scientist at AI payments firm Previse. “The key is that AI needs to enable a clear win-win situation which ensures that all parties see that they benefit from an AI-based solution that was not possible before,” he tells GTNews.

Can you trust an algorithm?

An issue of contention is the type of criteria is being used to make decisions and whether it flouts fair lending laws and regulations. Without being able to predict how a system is going to use the data, businesses need to be extra vigilant with the type of data being input. They should also be able to explain why the decision was reached, argues Cocagne.

Hodgson says: “Potent algorithms for AI and machine learning (ML) are already widely available in the finance industry. The problem isn’t so much about having the latest and greatest AI/ML – it’s about the lack of the right sort of data to power the algorithms in a consistent format, accessible in one place.”

“Due to the lack of clean data, a typical time split for a data scientist/analyst in a large bank with multiple data stores and formats is probably 90% on just data transformation and only 10% collectively for all the much more valuable work, such as selecting the right algorithms to apply to the data,” he suggests.

This means that banks are only able to draw limited insights from the data, and AI is only as good as the data which powers it. “Once institutions get their data capture systems right, we will see a real explosion in the power of AI, and trust in its capabilities will follow quickly,” argues Hodgson.

The lack of publicly-available data also means the technology is typically only available to business insiders. However, Shoenbucher says that this group normally lacks the expertise in machine learning. “Plus they are not focused on an application that is in the domain of external providers like supply chain finance,” he tells GTNews.

Bank-generated internal data is where some of the insights with real competitive advantages are held; however, this is generally underutilised. Businesses are spending significant amounts on buying market data from trading venues and exchanges, says Hodgson. “Market data is available to anyone with deep enough pockets, but no other institutions have access to this internal data,” he adds.

Many interactions in financial services still happen using voice or unstructured chat, which is hard to systematically access and analyse. Advances in AI allow businesses to capture this data in real time, and use it to give the best possible support to sales and traders, and better insight to management, Hodgson believes.

“One of the big opportunities [for AI] is to help banks to more effectively draw insights out of the data they already hold and then share those insights widely throughout the organisation. It isn’t just the quantitative analysts and senior management who can benefit from AI insights, but everyone from the sales teams to the risk managers and beyond,” he says.

“A key step in the technology’s development will be predictive analytics, so banks can receive specific action points, rather than just insight from AI – for example, adjusting inventory accurately based on predicting shifts in buy-side demand to capture more business.
“Analytics, including advanced machine learning and AI, are being explored by banks to integrate risk considerations (market, positional, counterparty), inventory, wider market factors and flow understanding,” adds Hodgson.

The technology is already finding its role in loan orientations as a sales adviser by helping with customer’s make decisions through the automatic analysis of supporting documents.

“It’s only by improving learning methods we can truly realise this technology’s real potential”

“It’s a great system to speed up processes, but on the flipside it boils down to whether people can trust the technology enough to allow an algorithm to manage their money, or if traditional human interaction will prevail,” says Cocagne.

“For now, it’s a combination of humans interpreting data presented by AI, including key analysis that helps to direct both sales and credit activities,” he adds.

What are the barriers to entry?

Acquiring an empty AI unit is relatively simple and gathering data from the back office is becoming easier too. However, the cost of AI systems is unpredictable and there are relatively high resource demands of in-house AI technology projects, says Hodgson.

An example of this would be training an AI system to function within a business. Cocagne believes this the most difficult task for any company. “It’s only by improving learning methods we can truly realise this technology’s real potential,” he says.

Hodgson adds: “Another frequent problem with building infrastructure in-house is that, if the IT staff in charge of the project are brilliant enough to make it succeed, they tend to move on to other things in a couple of years – thus, partnering with an external provider will give a bank much better continuity of support and development of the product.”

For this reason, traditional financial services partnering with fintechs can be very effective as fintechs have already covered the cost and resources of building technology and have the resources to efficiently support and enhance it. “By opting to outsource trading technology and paying attention to their core business, institutions can quickly deploy cutting-edge systems,” says Hodgson.