OCBC Bank brings audit to the digital age by harnessing Artificial Intelligence
Singapore, 11 September 2018 – OCBC Bank has piloted two financial technology (Fintech) solutions that leverage artificial intelligence (AI) to bolster its internal controls to safeguard the interests of customers and shareholders. The AI solutions augment the bank’s competency in the detection of trading anomalies during the audit of trading activities. This is done by analysing trade data using machine learning algorithms.
The two Fintech companies – Scila and Cardabel – were among eight that were shortlisted to be part of the 2018 TOV Innovation Challenge.
The annual challenge is one of the programmes run by OCBC Bank’s Fintech unit, The Open Vault at OCBC (TOV), aimed at providing innovative solutions to enhance customer-facing and internal processes.
Since TOV’s inception in 2016, it has engaged over 1,600 Fintech companies. Some of the digital tools that have been rolled out as a result of the engagements include OCBC Bank’s AI-powered home and renovation loan chatbot service Emma and a transaction monitoring system which utilises AI and machine learning to combat money laundering.
Detection of Trade Anomalies using Artificial Intelligence
The monitoring of trading activities to detect potential trade anomalies is currently carried out on multiple fronts by the Bank. These include daily checks by risk and control units and, as a last line of defence, by trained internal auditors.
Today, it can take an auditor up to three months to complete an audit of trading activities. The process involves considerable time to extract and sort voluminous trade and market data before manually analysing them to detect potential irregularities based on a set of pre-determined indicators. Thereafter, the auditor will investigate the irregularities.
It is ideal to deploy AI solutions for the auditing process as trading activities generate large volumes of digitised data. The following benefits are expected to be reaped:
How the AI Solutions Work
Scila (Sweden) has developed more than 100 market abuse indicators – a preset list of known anomalies such as insider trading – that their clients can leverage. Exceptions flagged and investigated can be fed into their supervised machine learning system to refine the market abuse parameters so as to improve the detection of "true positives".
Cardabel (France) uses unsupervised machine learning to detect both known and unknown types of trade anomalies. Its algorithms do not require preset rules to look for unusual trade patterns that have not been identified previously.
Ms Goh Chin Yee, OCBC Bank’s Head of Group Audit, said: "As new risk trends and anomalies continue to emerge for activities in the dynamic global markets, there is a pressing need for us to proactively and accurately identify and respond to them in an efficient and effective way.
"AI has shown an ability to not just analyse huge volumes of data and generate meaningful insights but be a powerful tool in identifying the unknown-unknowns in trade anomalies. Through AI, we will be able to further augment our audit effectiveness."
Six other Fintech Pilot Projects Completed
The 2018 TOV Innovation Challenge, now in its third year, saw the completion of eight Fintech projects over a period of 12 weeks. This is the first year the programme was extended outside of Singapore to two business units within OCBC Bank (Malaysia) Berhad.
Mr Pranav Seth, OCBC Bank’s Head of E-Business, Business Transformation and Fintech and Innovation Group, said: "The Bank is undergoing a pervasive digital transformation across its businesses, geographies and strategic functions to change the way it operates. The Open Vault at OCBC has focused its efforts to do this through an energised culture of experimentation and learning.
"While the Challenge has allowed us to find ways to improve our internal capabilities with artificial intelligence to bring greater value to our customers, it is just the tip of the iceberg. The broader vision for us is to see to fruition the transformation of every element of the business from customer-facing operations to back-end processes."
Besides the pilot tests with Scila and Cardabel, OCBC Bank also completed pilot tests that leverage AI with six other Fintech companies as part of the programme.
Icekredit (China) leverages big data and machine learning to help create strategies that are predictive and personalised to collect repayments, based on a customer’s personal and professional background. The solution helps to identify high-risk customers through a scorecard and then suggests the best method to engage them.
Squirro (Switzerland) leverages predictive analysis and machine learning to help relationship managers turn unstructured digital data into actionable insights for their customers in real time.
Lang.ai (Spain) analyses text converted from call audio files received by OCBC Bank’s Contact Centre by using unsupervised AI to track the conduct of both staff and callers in customer service engagements, cross-selling attempts and compliance management. Contact Centre staff currently listens to and reviews a sample of about 1,200 of the 150,000 customer calls it receives every month.
vPhrase (India) automatically summarises data from various financial sources into key insights for data-driven decisions to help relationship managers who serve small and medium businesses to create credit proposals and review statements. This tool can reduce the time taken to create these documents by 30%.
EZMCOM (USA) provides a completely digital customer experience with instant verification of business documents, certificates, persons and IDs. Verifications include behavioural authentication and fusion biometrics.
Ondot (USA) provides a cardholder with a loyalty programme, contextualised reward and spend insights, secured card control features as well as a social aspect for the sharing of news.