Continuous self-learning cuts fraud and risk
Question: What have been the qualitative and quantitative improvements in risk management and fraud control systems by deploying AI?
Anjana Rao is Chief Strategy Officer at IndiaFirst Life Insurance
IndiaFirst Life has rolled out a risk score model, which predicts the propensity to fraud. We plan to make it a part of our onboarding journey to prevent such applications from being logged or to push such applications into a higher due diligence bucket. A model to predict early claims too has been deployed. AI will play a significant role in strengthening the risk writing process of the organization. This space is yet to evolve.
Girish Nayak is Chief - Customer Service, Technology and Operations at ICICI Lombard General Insurance
Risk management is the core of any insurance organization and we continue to see new technologies helping in overall risk management. On the commercial i nsurance s i de, we hel p pr o v i de customized solutions to our customers that focus on customer-specific risks. IoT and drones are playing a significant role in this aspect. We use several IoT device-based solutions that help our customers in hydrant monitoring, firefighting equipment monitoring and in electrical risk solutions. Similarly, for our marine cargo customers, edge-based IoT telematics devices are helping in providing tracking basis solutions and instant event-based alerts.
If you think about the current covid vaccines, it requires transportation using a cold supply chain and refrigerated cargo. For pharmacy, temperature control is very essential for transportation of these vaccines. In India at times since drivers are incentivized to conserve fuel, they are likely to switch off the refrigeration units for some time. We immediately get a trigger of this event at the time the temperature increases or decreases basis on thresholds that are set in these IoT devices. This helps us to save cargo from being damaged.
Similarly, with one of our large metal company clients, we have built a just in time cargo solution using triangulation of the IoT device data as well as the mobile data that is available. There are multiple such examples where we have been able to help our clients mitigate multiple types of risks such as hijacks, thef t and pilferages, temperature excursions in pharmaceutical items and moisture sensitive cargos. Sophisticated algorithms help us in analyzing the vast amount of data that is generated through these IoT devices and help us on improving solutions for our customers. On the retail side, IoT based telematics devices in private cars are helping us in developing driving behavior-based segmentation models.
On the claims side, we have been developing AI/ML solutions such as our cashless pre-authorization for group health claims where an algorithm decides the admissibility and authorization amount for a particular surgery/illness within 90 seconds. However, to ensure this seamless service, we needed to create solutions that help us in identifying those customers who can potentially take undue advantage of the system. So, each one of our digital claims solutions is backed by powerful fraud identification solutions that help us in identifying new patterns of fraud that keep emerging.
Earlier, we would flag off claims that needed scrutiny basis rules and triggers that were put in the system. This would be investigated thoroughly and marked as fraudulent or kosher basis the investigation. Today, with significant data availability and technology, we have started using AI and ML based fraud detection models to predict and highlight probable fraudulent claims in real-time. All this happens immediately, once the claim is intimated and logged within the system. The continuous self-learning approach of these ML algorithms helps us to implement solutions that autocorrect very quickly reducing the time for learning and execution. These algorithms are helping us in identifying and prevent frauds faster and giving us higher efficiency.
KV Dipu is Head – Operations & Customer Service at Bajaj Allianz General Insurance
AI is more useful when there is huge data and hence deploying AI for risk management and fraud control becomes a breeze. Learning from past patterns and incidents and converting them into algorithms (AI) coupled with HI (Human Intelligence) has yielded both qualitative and quantitative results for us.
Supreet Badrinath is Vice President - Business Development Data & Services, South Asia at Mastercard
Risk related solutions are managed by our Cyber and Intelligence Division. Mastercard has grown its capabilities through acquisition of companies such as Brighterion, which brings in unique AI capabilities towards prevention of payment and acquirer fraud. Another such acquisition is RiskRecon, which deploys AI to help businesses and financial institutions proactively manage cyber risks, safeguard critical intellectual property and consumer and payment data.
Udayan Joshi is President of Claims & Personal Lines Underwriting at Liberty General Insurance
AI deployment has augmented our digitization journey. AI algorithms work logically without emotions, making rational decisions with lesser mistakes and promotes automation in the process chain. It also helps us in detecting frauds and leakages such as detection of prior damages and hidden damages. In medium to long term, it produces a lower error rate compared to humans. By taking over repetitive and tedious tasks, it offers us an opportunity to use our talent more effectively in other areas of business.