BFSI Accelerator 2022

Unlocking the Future for Banking, Financial Services and Insurance (BFSI) Industries


A1. Financial Wellbeing
Scenario (1)
“The bank is currently faced with the issue of providing a service to our customers that assists their financial wellbeing, and also improves our relationship with them. We currently do not have a service that accumulates all customer essential costs and bills. This includes expenses such as utilities, rent, mortgage, education, transport and more.”
#DataIntelligence #OCR #NLP #ReportingAndAnalytics #FinancialTracking

A2. Footprint Tracking / Evaluation
Scenario (1)
“We are relying more on digital channels to not only collect customers’ online footprint, but also for more efficient and effective methodology to analyse the data. Our issues include accumulating a huge volume of data and also incomplete data due to web tagging limitations. We also lack an efficient approach to data analysis and workflow, and don’t have an appropriate algorithm model for sequence identification. This flows into our inability to provide a personalised customer digital journey from onboarding to transaction.”
#DataAnalytics #WebMonitoring #DigitalFootprint
Scenario (2)
“Tagging requirements are essential when launching new features on web and mobile - both at page and tab level. We would like to explore the benefits of automatic field tagging tools for customer service, as omissions of tagging can lead to limitations on result tracking. This makes it very difficult for businesses to review performance and limits follow-up action as businesses do not know where customers drop off. We are interested in the ability to track missing tags at tab level, as well as a summarised library to show all the information architecture with relevant tagging.”
#WebTagging #WebArchitecture #Automation

A3. Knowledge Base
Scenario (1)
“We don’t have a comprehensive and centralised platform to integrate client profiles - including sales activity, communications, credit situation and suitability. This means customer reports and data are not stored efficiently for later use. Relationship managers end up spending lots of time to find the correct data, and it’s also difficult for them to ensure the real-time performance of data. It proves unhelpful for them as they cannot retrieve accurate information between the previous relationship manager and the customer. This means they cannot support clients’ timely needs and detect potential risks early.”
#DataIntelligence #DataVisualisation
Scenario (2)
“It requires plenty of time and manual effort to collect company data that is open to the public, for example from the Hong Kong Companies Registry. There is also a lack of access to useful data including information of newly incorporated companies. While many third-party vendors provide useful customer data, their database usually contains corporate-sized enterprises and not SMEs. Banks also lack the scalable data aggregating capability to quickly connect to data providers, and the data quality isn’t compatible with banks’ requirements. This complicates procedures such as credit lending approvals, customer onboarding and customer due diligence.”
#DataIntelligence #API #AlternativeData
Scenario (3)
“We share information across different SaaS products within a business process, where we have to download or extract from one provider before uploading it onto another. This inefficient method increases the risk of human error. We also lack a service provider that can take data across end-to-end business processes. Data for our business is unorganised and scattered across many service providers, so we don’t have a holistic view on performance or projections. It is our aim to provide data aggregator capabilities across multiple SaaS software.”
#DataAggregation #SaaS

A4. News and Trend Insights
Scenario (1)
“Insurance companies don’t have the appropriate data or assessment method for their customers. They need more insights behind just the disease names or statistics on medical reports. Without trends showing the development of applicants’ health, it is difficult to set up specialised health schemes that help customers to fight against diseases together. Customers are not being provided with healthy advice as we cannot uncover hidden health indicators. We would like to disclose any hidden health indicators to customers, so we can contribute to providing health advice to society.”
#DataAnalytics #AI
Scenario (2)
“The marketing and events team spends significant resources to polish its wordings, while other business team members rely heavily on Excel and PowerPoint reporting. Many support staff spend much of their time dedicated to reporting work so they don’t have the time to carefully craft keywords. Also, natural language processing is steadily becoming a widely adopted data solution, and it is proving to be successful in assisting and facilitating operations. We require greater keyword training and sentence recommendations for staff when they’re producing marketing materials.”
#SocialListening #NLP #KeywordTraining
Scenario (3)
“Our biggest problem is information collection, which takes up most of our resources. It takes a lot of time to approach information providers, and then to filter messages and clean the data before we can see the insights, costing us time and money. However, we need this critical input into our product design process - including retirement financial needs, customer trends and medical service systems. We’re collating all this data from a variety of media including government published statistical reports, research institutions and others. This is not an efficient or cost-effective approach.”
#DataCollection #SocialListening #NLP #DataAnalytics

A5. NFTs, Web3 and the Metaverse
Scenario (1)
“The NFT, Web3 and Metaverse markets are heating up and we are interested to learn about the market perspective on potential opportunities for banking and financial services. How can we turn the hype into business opportunities?”
#NFTs #Web3 #Metaverse

A6. Precision Targeting
Scenario (1)
“When attempting to target high net worth customers, we find it difficult to collect useful third-party data to identify their assets under management (AUM). In an ideal world, we’d like to accurately map customers’ wealth to service them with more targeted products. But with limited internal customer behavioural data, we are unable to understand customers’ product and channel preferences, and difficult to visualise their total wealth.”
#AlternativeData #DataIntelligence #Visualisation
Scenario (2)
“Our business is experiencing a loss of opportunities when trying to identify customers’ needs. We’re unable to deep dive and offer targeted banking solutions to increase the success rate. Currently, we use different channels to contact customers for various products and services - we do not have a centralised platform or database to collect responses and feedback. Customers might even visit branches to complete transactions over the counters, and even use our website for information, but these customer insights are not tracked. This makes the job for relationship managers even more challenging.
#O2O #BehaviouralTracking #Personalisation #DataTracking
Scenario (3)
“A digital-led acquisition strategy for the life insurance sector is not a new phenomenon. Most new leads come from digital campaigns, as opposed to cold searches or referrals in traditional methods. But due to the complexity of insurance contracts, most of the touch points to clients and potential customers are through agent engagement. We would like to provide better agent to client matching capabilities and personalised services, through the provision of alternative data about the clients and the categorisation of prospective clients. That’s why we need a digital-enabled insurance sales methodology and explore the potential of automated client profiling and client matching mechanisms, which will allow us to better service our customers as well as attract new ones”
#AgentToClient #Personalisation #AlternativeData #ClientSegmentation
Scenario (4)
“We currently lack customer experience-related processes, including analytics and digital tools/platforms to manage the entire banking customer journey. This gap in our capabilities ranges from awareness, consideration, purchase, retention and advocacy, which are required to properly identify customer pain points or issues. This is especially apparent in digital channels such as mobile and online banking, which are crucial to achieve customer acquisition and retention. We are also lacking in customer insights, including what preferences and behaviour they have towards types of financial products. Nowadays, banking customers expect tailor-made products but our internal data makes it limited to acquire more customer information for enhanced offerings.”
#AppliedAI #MarketingIntelligence #Geofencing #UX #UI


B1. Early Warnings
Scenario (1)
“Banks can only intervene to help our customers to investigate fraud cases after they have been reported to us - by that time their transaction has already been processed or legal action is required from police. There is a lack of a consolidated database to store and record fraud cases, with critical information and details of related parties. Banks also lack the required tools to process the data and instantly screen transaction details against the database. This restricts their ability to provide proactive advice to SMEs about fraud cases.”
#DataIntelligence #AlternativeData #FraudDetection
Scenario (2)
“Enquiries for suspicious transaction details are triggering an overwhelming number of calls to banks. Without real-time fraud alerts, our frontline staff are unable to take preventative and proactive measures to alert customers. And agents are spending an unnecessary amount of time on false positive calls, which they should be focusing on high-risk alerts. This often prolongs the investigation process, where customers then have to provide evidence. This is a bad customer experience and can lead to financial loss for the customer and the bank.”
#AppliedAI #TrendsAndInsights #CustomerExperience #DigitalJourney
Scenario (3)
“The bank relies on historical data and professional experience to effectively plan capacity, the number of staff members on duty and which departments they will focus on. For the team manager, they have limited visibility of this data and they usually must perform analysis based on their judgement and best effort. We have the issue of not having the data relating to fraud alerts peak times, common fraud types and more. We could benefit from a dashboard that presents recent fraud trends as it will give insights for staff capacity planning.”
#DataVisualisation #WorkforcePlanning #DataAnalysis

B2. Fraud & Mule Account Detection and Visualisation
Scenario (1)
“Virtual banks can lean on the experience of other mature banks to identify default clients, or fraud and money laundering clients. With the continued expansion and development of our bank, this requires more investigation into anti-fraud methods. This strategy should comply with local laws and regulation and be adaptive to Big Data and AI technology to avoid greater exploitation. We would like to explore previous strategies that we can adapt to our bank, including using AI technology and Big Data. This would be more effective in identifying default clients without disturbing well-performing customers.”
#AlternativeData #AppliedAI #FraudDetection #AntiMoneyLaundering
Scenario (2)
“Although we have implemented a rules-based detection system to identify abnormal account activities, the adoption of Open Source tools (such as RStudio/Python2) for network detection and visualisation has created some new challenges. We are still encountering issues with visualising identified potential connections with complex, multi-dimensional data. This makes the presentation diagram unorganised and messy - there are too many attributes and entities included. This requires more resources for redrawing if more observations are unearthed during analysis. There is a need for a responsive and interactive network diagram tool.”
#DataVisualisation #DataProcessing #NetworkDetection

B3. Impersonate Risk
Scenario (1)
“Banks have traditional log-in security safe measures such as password, soft token and device binding, and SMS OTP. Virtual banks can adopt a new authentication technique using facial recognition. But traditional banks are unable to require customers to redo facial recognition after their account has been opened, especially after they move overseas where SMS OTP is no longer an option as they have changed their registered mobile phone number. We recognise that while SMS OTP serves as a factor of authentication, it is becoming increasingly prone to fraud, with criminals able to bypass the authentication.”
#Cybersecurity #Authentication #FacialRecognition #VideoAnalytics #AppliedAI
Scenario (2)
“In the health insurance industry, there can be situations where there is insufficient information to conduct a thorough risk assessment on a person’s health (including their history and trends), and then relate them to underwriting, monitoring and customer service. This opens up the risk of fraud. The customers’ endorsement to collect health information for underwriting and regulation is also a known issue. Due to personal data sensitivity, many prefer not to disclose this to external parties, which makes it challenging for us to understand the entire profile of customers’ health in the wider market. We have issues in properly and legally collecting consent from customers. In health monitoring, we also have issues in ensuring the readings are coming from the right person when the incentive to commit fraud is large.”
#IOT #Authentication #DataSecurity #PersonalData

B4. Staff Data Leakage
Scenario (1)
“The bank is facing an increased risk of data leakage as staff members are accessing the customer database outside of office premises. Becoming more prominent during the work from home policy introduced since the COVID-19 pandemic, this includes staff taking photos of customer data and then sharing it with a third party, heightening the risk of outside interference. We want to explore the technology options available to mitigate this data leakage risk and improve our response to data security and any potential breaches.”
#Cybersecurity #RiskAssessment #LocationBased #DataSecurity

B5. Transaction Monitoring
Scenario (1)
“Credit card issuers are always developing new methods to stop unauthorised transactions, as cyber criminals continue to search for new ways to combat us. Our necessary investigation periods can last up to one month, which means our customers are not being informed in time by credit card companies. This leads to them being unable to take any preventative measures before suspicious transactions occur, including disabling online transactions of “card-not-present” - a built-in feature of mobile apps. This leads to a poor customer journey and experience.”
#ScamDetection #LocationBased #DataIntelligence #CustomerJourney

B6. Detecting Potential Fraud Signals from Unstructured Data
Scenario (1)
“There is improvement to be made in fraud detection and the screening process by insurer claims examiners. Usually, our data is provided by insurers or underlying insurers in an unstructured format, and comprising multiple claims, payment screens, copies of invoices and reports in various languages in the Asia-Pacific region. We would highly benefit from AI-led, intelligent screening to identify potential red flags, including the same invoice being used to support multiple claims. This will increase the chances of detecting potential fraud or red flags." 
#AppliedAI #UnstructuredDataProcessing #OCR #NLP #IntelligentScreening


C1. Blacklist Management
Scenario (1)
“Blacklists are crucial in financial services, particularly banks, but these institutions can painstakingly spend copious amounts of time to add a single customer to a blacklist. By manually adding entries to an Excel spreadsheet is not only time-consuming - this manual input can result in human error. While mistakes are part of human nature, this can pose a significant anti-money laundering risk to the bank if a blacklisted person is successfully onboarded. We also come across a lack of a centralised system which creates more inefficiencies. For example, we have two separate systems when blocking an onboarding customer and also blocking an application.”
#DataProcessing #RPA #AntiMoneyLaundering

C2. Due Diligence
Scenario (1)
“According to anti-money laundering guidelines, banks must complete ongoing customer due diligence. The bank has invested in a workflow system to review documents, data and information to ensure they’re updated. But this can be time consuming to obtain, especially with corporate and high-risk customers as many do not respond to requests for updated information. Also, our customer information is stored across different systems, and we lack a centralised panel for a holistic review on customer profiles. This hampers our ability to conduct due diligence, as we don’t have access to the latest static information including data collected during account opening, enhanced due diligence, periodic review, information collected during loan application and transaction monitoring alerts.”
#DataVisualisation #DataProcessing #NewsScraping #AppliedAI #AntiMoneyLaundering
Scenario (2)
“Customer Due Diligence (CDD) is one of the key elements of financial crime risk and anti-money laundering control. For clients with higher financial crime risk, Enhanced Due Diligence (EDD) is required, in which the bank has to understand how the client accumulated their source of wealth (SoW) with a detailed summary documented in the client profile. We require more information into SoW summaries with more detailed information in client profiles. We would like to know if there is opportunity for an end-to-end solution that leads to a more efficient SoW summary.”
#DataProcessing #AutomatedReportGeneration #ReportingAnalytics
Scenario (3)
“Open source news search is a part of the customer or transaction due diligence process, especially in high-risk situations. This is currently completed manually on search engines including Google and Baidu. Investigators have to sort through a large number of search results with many false hits - a time-consuming process. Then, they must spend more time to understand the context of the news to identify key topics and determine if the news is actually related to the subject. There are also different challenges when dealing with Chinese and English news.”
#NewsScraping #AppliedAI #OnlineMedia #SearchEngine

C3. Name Screening
Scenario (1)
“As name screening is an integral part of the bank’s work in combatting money laundering and terrorism financing, this requires a lot of time and manual effort to investigate and process payments and customer name screening alerts. Many false positives are generated by the name screening engine and need resources for clearance. As all alerts require manual handling, this slows down the overall speed and efficiency of transaction processing and opening of accounts. This will only become more problematic as the business grows.”
#NewsScraping #AppliedAI #NLP #DataVisualisation
Scenario (2)
“We have various parties including exchange, clearing and investors who must undergo the know your customer and due diligence process for UN Sanction screening. Currently, a heavy manual support or self-developed program is required to conduct due diligence, and sanction list screening data (news, search function, websites) are mostly scattered. The current system also does not provide timely updates on sanctioned entities, so we have to use the search function to check. We are currently exploring other sanction screening platforms to be adopted throughout the firm.”
#DataIntelligence #NLP #AppliedAI #RPA #DataVisualisation

C4. Suspicious Transaction Reporting
Scenario (1)
“Our team invests significant resources to manually fill in suspicious transaction (STR) report forms, by copying and pasting customers’ static information from different systems. These include name, ID number, address, email, mobile number and account numbers. This archaic system means an employee has to verify the same information, which leads to human error in typos and missing the blanks. Other banks have implemented automatic extraction of this data to the STR for highest efficiency and accuracy. We would benefit from software to collect data from various systems.”
#UnstructuredData #ReportingAnalytics #NLP #RPA #OCR

C5. Transaction Monitoring
Scenario (1)
“Our current remittance payment system is screened by the FX Anti-Money Laundering (FFM) system to identify messages that contain predefined sensitive keywords. This system generates more than 50,000 alerted messages per month, which are manually checked. The bank experiences a high number of false positive alerts which has eroded customer trust and critical inefficiencies when comparing the time spent against actual financial losses. The bank’s reputation has taken a negative hit and we are spending too many resources in dealing with false positives. We also have difficulties with data privacy and security, with banks and other financial institutions not permitted to share their payment data, making it challenging to find fraud patterns.”
#DataAggregator #RealTimeEngine #NewsScraping #AppliedAI #AntiMoneyLaundering
Scenario (2)
“Anti-money laundering (AML) practices are essential to enhance the stability of the financial system and profitability of authorised institutions. While we place the utmost importance on AML strategies, our current system and transaction monitoring tool are rule-based, and not optimised for detection accuracy and human intervention. We are dedicating numerous staff from multiple departments to spend long hours to handle and check false positive alerts. We are also hit by the constant challenge of adequately handling external data from third party vendors, other financial institutions and market information. This data and insight is not leveraged and correlated to detect AML patterns.”
#DataIntelligence #API #AlternativeData
Scenario (3)
“Banking teams do not possess the full picture when handling banking alerts, with a lack of full visibility on all customer activity including log-on, before and after image of contact changes and adding a new beneficiary. Agents instead have to use their best judgement and assumption when assessing alerts, or else they must spend considerable time to navigate different systems to uncover necessary info. This prolongs the handling time of each alert, and determining the risk level. We need solutions that offer quick assessment and decision making for agents.” 
#DataVisualisation #RiskAssessment


D1. Alternative Risk Assessment / Risk-Based Pricing
Scenario (1)
“Credit decisions form many pain points for financial institutions. Traditional underwriting systems make critical decisions using few data points, heavily relying on credit bureaus. This often results in people with limited credit history being denied credit, which negatively impacts banks (loss of revenue) with a smaller pool of potential customers. There are also challenges for wealthier customers, with inefficiencies and red tape extending application times - even for existing customers. Also for small and medium enterprises, their lack of audited financial statements provide a barrier for lenders and their ability to make calculated lending decisions.”
#DataIntelligence #AlternativeData
Scenario (2)
“Banks have a greater need to gather and incorporate alternative and ESG (Environmental, Social and Governance) data to enrich and enhance their credit scoring models. This means more quantitative data sources - such as the tracking of a corporate’s carbon footprint - are required to enable banks to execute more accurate credit decisions for assessing green loans in corporate lending. By increasing these alternative data sources, this will contribute greater interoperability with the HKMA-led Commercial Data Interchange data and ESG data for improved underwriting.”
#DataIntelligence #AlternativeData #ESG #CDI #GreenLoan
Scenario (3)
“As a virtual bank, we are lacking in the capability of obtaining alternative data and more advanced data sources. Our current data sources are quite limited and we need to improve in the gathering of more useful data to upgrade our credit scoring mechanics and credit decisioning model. Ultimately, we seek the development of a multi-dimension credit model for greater data intelligence and capture, that will allow us to provide better services to our customers, including convenience and flexibility.”
#CreditModel #DataIntelligence #AlternativeData #BigData

D2. Compliance Risk - Customer Endorsement
Scenario (1)
Banks face difficulty in obtaining consent from customers to collect their personal data, especially alternative data. This includes valuable data for banks such as credit card transactions, smartphone activity, social media sentiment, IoT sensor, app usage and satellite imagery. Due to data sensitivity and privacy concerns, many customers prefer not to release their personal data. We want to properly and legally get consent from customers to collect and use such alternative data to guide our business.
#DataCollection #DataSecurity #PrivacyRisk #CustomerConsent
Scenario (2)
“There is a significant upside to generating value and returns for customers and the bank, by sharing aggregate customer data with third-party vendors. But this is often an insurmountable obstacle due to data security and privacy concerns that restrict banks from sharing data which leads to the question - how can we manage customers’ consent and share data with third parties, with the pay-off being rewards for those customers? At this moment, customers both do not have control of their own data, and do not use this data to their advantage, which also means banks are losing out.”
#AggregateData #DataSecurity #PrivacyRisk #Encryption 

D3. Credit Spread Market Information
Scenario (1)
“When some banks enter into repurchase or reverse repurchase transactions with clients, they often use bonds as the underlying collateral as it presents a medium of financing with lower-risk compared to a loan. However this brings issues including the lack of information source to track the credit spread movements of bonds, as well as the issuer of bonds. It also makes it difficult for the repurchase department to accurately and quickly review the credit profile of the collateral portfolio. Some analysis is available through credit research firms but this can prove costly for banks.”
#DataIntelligence #CreditSpread #Tracking

D4. Legal Risk
Scenario (1)
“In the legal sector, the Country Addendum for a specific Group policy or standard usually contains a list of local laws and regulations that often constitute variations from Group requirements. What transpires are interpretation and operational guidelines being requested by stakeholders, with this communication repetitive, inconsistent and inefficient. Frontline staff members may not have the knowledge or experience to take the required actions, while various stakeholders may have different tasks within the Country Addendum but are forced to look up the document in its entirety.”
#NLP #AppliedAI #StakeholderEngagement