MACHINE LEARNING IN TRADE FINANCE: WHO’S DOING WHAT? by Shannon Manders, ITFA Consultant



The ITFA Fintech Event in June, kindly hosted by Deutsche Bank in London, uncovered different ways in which data management and machine learning technologies are being applied to trade finance origination and distribution.

The event profiled the transformational work being undertaken by the likes of Taulia (corporate working capital), Traydstream (trade finance documentation), Coriolis (trade insights), Tradeteq(transaction-level credit scoring) and INTIX (data management). A video of the entire event can be watched here, but these are some of the highlights from the discussion.

  1. Regulators want the banking industry to innovate

“Regulators are taking a keen interest in how banks are deploying new technologies: they want banks to be able to compete with the fintechs and offer better service to clients.” This was one of the messages delivered by Polina Evstifeeva, head of regulatory strategy, GTB Digital at Deutsche Bank, whose remitsees her looking at regulation in the digital space and actively engaging with regulators on the different rules shaping the industry.

Speaking at the ITFA event, Evstifeeva delivered a presentation which outlined all of the current digital regulatory trends and highlighted the various instances where regulators have thrown their weight behind supporting innovation in banking. One of the key developments from last year, she said, was the European Banking Authority’s guidance on outsourcing, including outsourcing to the cloud services providers, which has helped to clarify rules around cloud computing, something which all banks are looking to employ.

Underpinning all of the new technologies changing the business of banking – whether it’s blockchain, APIs, cloud or artificial intelligence – is data, one of the modern economy’s most valuable commodities. The security of this data has always been of interest to regulators, and now they are turning their attention to the ethical use of client data, Evstifeevasaid. “This includeswhat you can use your clients’ data for.” To this end, the European Commission High-Level Expert Group (EC HLEG) hasrecently developed non-mandatory Ethics Guidelines for Trustworthy AI, such as the need for transparency, explainability, liability and human oversight when using client data, which may go on to shape the future policy agenda. According to Evstifeeva, these guidelines will also go through a piloting process, expected to conclude with the presentation of a revised document in early 2020.

A Deutsche Bank whitepaper entitled “Regulation driving banking transformation” released in October 2018, applauds the regulatory efforts to date to balance innovation and risk management, but outlines areas for improvement, namely the need for acceptance of the new realities created by emerging technologies, and the need for further global regulatory alignment.

“Banks are re-imagining their operations, processes and solutions in this new, digital data world. It is important that regulators keep pace with such change,” the paper reads. It goes on to highlight a number of areas where issues are being faced, such as, for example, the European Union’s General Data Protection Regulation (GDPR), which enshrines the “right to be forgotten”, but potentially hinders the opportunities derived from the immutability of blockchain. “We are on the cusp of major change. And forward-thinking regulation stands to be a major catalyst for a thriving and innovative banking industry.”

  1. It’s time to embrace the data-centric shift

“The future of trade is data,” said Rebecca Harding, CEO of Coriolis Technologies, as she introduced her company’s technology, MultiLateral, to the audience.

“Our technology connects the dots between disparate data sets. If we can understand the data flows, we can understand what governments, businesses and banksare doing – and then we can understand trade,” Harding said. She described MultiLateral as the “Bloomberg” for trade finance in that it provides a global and macro picture of what’s going on in the world of trade, which parties can then use to make decisions around financing it.

“Over the last decade, various industries – in particular retail – have shifted from being software-centric to becoming data-centric,” said André Casterman, CMO of INTIX, a data management fintech, and chair of ITFA’s Fintech Committee. “The trade finance market is starting to follow this trend too as machine learning is entering the space to further improve business practices. Accessing data, analysing it with machine learning and delivering continuous business insights is increasingly proving its value to trade bankers.”

Coriolis not only aggregates the data, but also uses artificial intelligence (AI), specifically machine learning, to create unique data analytics for its clients, which include banks, investors and policy makers. Among other things it provides sentiment analysis to identify company, counterparty and compliance risks.

  1. Machine learning has many applications in trade origination and distribution

This method of complex data analysis utilising machine learning is also carried out by the likes of Taulia, Traydstream and Tradeteq.

Taulia, as Christophe Juvanon, head of solutions consulting EMEA, explained at the event, delivers working capital solutions that make it easy for businesses to free up cash, accelerate payments and improve supply chain health.

Taulia runs a buyer-led platform, now on its third generation, which uses AI capability, meaning that buyers can better assess supplier behaviour and risk.

Simply put, the platform makes it effortless for corporate buyers to liberate cash within the supply chain, Juvanon said.

Taking this one step further, in April Taulia partnered with Google Cloud to launch an AI-powered invoicing solution. Called Cognitive Invoicing™, the solution enables businesses to successfully process invoices in any file format from their suppliers. It leverages the optical character recognition (OCR) capabilities of Google Cloud’s Document Understanding AI to read and interpret unstructured supplier invoice data in Taulia’s global invoicing platform. According to Taulia, it will reduce the cost and time associated with processing an invoice.

Traydstream’s solution, meanwhile, enables banks and corporates to automate the manual scrutiny of trade finance documentation. As explained by its CEO, former banker Sameer Sehgal, it combines machine learning and AI with Natural Language Processing (NLP) to create intelligent pattern recognition and near-instant verification of regulatory compliance.

In essence, it reads, scans and structures paper-based trade information digitally. It then uses AI to automatically process and check the documents for anti-money laundering and compliance issues – a job which, for many banks, remains alabour-intensive process.

Since Sehgal’s presentation at the ITFA event, Traydstream has announced a new pilot partnership with Finnish multinational exporter Nokia Corporation and several of its major advising banks. The 10-week pilot will see Nokia trial the platform with three of its key banks, OP Group, SEB and Standard Chartered Bank.

Tradeteq operates a cloud-based trade finance distribution platform. The company’s head of AI, Michael Boguslavsky, explained how it works: “Tradeteq provides securitisation as a service, transforming trade finance receivables into investable assets. We provide the network, connecting trade finance originators, funders and services. And we also provide better transparency for the market by providing company and transactional credit scoring,” he explained.

As of earlier this year, Tradeteq leads the Trade Finance Distribution (TFD) Initiative, a consortium of 19 global financial institutions, to create common data standards and definitions for global trade finance asset distribution.

Asked at the event about the future of machine learning, Boguslavsky responded that it can go in several directions, including improving models, many of which still require much work. But, he said, in many cases, the biggest progress will be achieved by putting more data into the models and changing business practices. Up-to-date data, taken from Internet of Things technology and data mining the blockchains to source the relevant trade information, will be essential and will enable new decision-making practices, he said.

  1. Specialised fintech-fintech collaboration is driving successful solutions

“Collaboration can come in two forms: either a bank can collaborate directly with a fintech, or – where the real magic is – is collaboration between the fintechs themselves,” said Daniel Rymer, head of FI trade solutions at Mizuho, and member of ITFA’s Fintech Committee.

Case in point: almost all of the fintechs that presented at the ITFA event are collaborating with INTIX, the management team of which André Casterman joined in 2016 as its chief marketing officer.

For the likes of Coriolis, Traydstream and Tradeteq, INTIX captures the transaction data, as well as master data, payments data, Swift messaging and so forth, and the related events, such as transaction AML screening, from internal back-office systems at granular level and in real time. This data is aggregated and passed on to the fintechs in a way that they require it in order to apply their specific value-added processing to it, and ultimately enable them to deliver the solution to the parties which have contracted them.

Casterman explained that machine learning needs electronic data – from the banking, trade or payments systems: data that is used day-to-day, but also data that dates back decades and is stored in archives. All of these systems can be plugged into the INTIX data management layer, which has the understanding of all the formats and data structures. “That creates the ‘information bus’ that is then fed to the AI-based entities of trade,” he said. “This makes the fintech-fintech collaboration quite nice in terms of combining very specialised roles and making sure that the client gets full control of the process. You keep the data in your own environment, there is no need to externalise it.”

He made the point that the aim of fintechs – which are often small and specialised, like the ones present at the event – is not to change what already works well. Instead, they’re a means of adding value where pain points are located, to drive even more use from existing systems.

“Rather than calling what exists today‘legacy systems’, which can sound negative, I like to call them ‘robust systems’. Because your back-office systems are robust, proven and the value add can come alongside them, rather than changing them,” Casterman concluded.