Digitalisation across the insurance value chain brings with it several benefits including improved customer experience, awareness, innovative products, and interactions with intermediaries. This has, however, led to increased exposure to fraud.
The PwC Global Economic Crime Survey 2022 records 63 percent of Eastern African respondents reported having experienced fraud compared to 46 percent globally. The need for automated risk management and fraud control using Artificial Intelligence, Machine Learning, and Blockchain to strengthen fraud prevention and protection is thus more crucial than ever before.
According to the Association of Certified Fraud Examiners (ACFE), a fraud control activity is intended to prevent or detect fraud quickly. Automated fraud control activities established through monitoring and real-time or near-real-time alerts are critical to improving fraud detection and prevention capabilities while reducing operational costs.
The automation must envisage the possibility of fraud occurring at any stage of the insurance value chain: application, policy issuance, customer service, claims processing, payments, or exit.
Automation leverages capabilities in data analysis, pattern recognition, and anomaly detection to process large volumes of data. This data is obtained from sources such as financial transactions, customer profiles, and historical records. Automated solutions examine relationships between different data points to detect fraudulent behaviour.
Let’s explore some of the common fraud schemes that can be detected through automation.
To start with, automated system controls integrated with other third-party systems can help detect fictitious claims through comparison of claims data against other sources such as historical data, policy information, and public records from regulators and other institutions to identify inconsistencies.
Next, automated systems can collect multiple data points and help in detecting staged accidents or intentional injuries by analysing claim data against factors such as location, type of damage, medical reports, unique motor features, and the number of passengers involved to identify any suspicious patterns.
Another instance is whereby automated systems can process large amounts of data in real-time to help detect medical billing fraud by analysing claim data against industry benchmarks, standardisations like International Classification of Diseases (ICD) codes, provider billing patterns, and patient histories to detect inconsistencies.
Automated systems can also help by analysing policy data against external sources such as credit reports, Integrated Population Register Systems (IPRS), and public records to identify premium fraud, that is any discrepancies and potentially stolen identities.
During policy admission and premium payments, they can also compare customer data, such as Social Security numbers, addresses, and employment history, and analyse patterns of application submissions to detect multiple applications using the same or similar identities.
Additionally, automation can be used in analysing claim data against medical records, employee histories, and industry benchmarks to detect inconsistencies that result in workers compensation fraud.
Moreover, automated tools can monitor trading activities, transaction volumes, and frequency of account changes to detect abnormal patterns indicative of churning which refers to excessive trading or switching of investment products to generate commissions for brokers or agents.
Lastly, automated systems can detect irregularities such as consistently high returns, excessive fund transfers, or a lack of legitimate underlying investments. Ponzi schemes involve using funds from new investors to pay returns to earlier investors, famously known as taking from Paul to pay Peter.
The lack of data, and good data quality because of errors or omissions, that include missing, inaccurate, or inconsistent data across operational systems is a big challenge for insurers. The challenges directly affect the efficacy of the automated systems and technology tools being implemented.
The lack of data governance structures across business units, operations, and other functions creates challenges in aggregating and analysing data.
Data governance and good data quality practices are essential if insurance companies are to effectively implement automated fraud risk systems. Data governance establishes the rules and processes for managing data across the organisation, ensuring that data used in fraud risk automation is reliable, accurate, and consistent, reducing the risk of errors and false positives.
It can also ensure that insurers comply with regulations, identify the right data sources, build a collaborative culture, and improve operational efficiency by continuously monitoring and improving systems.
Automating fraud risk management using Machine Learning, Artificial Intelligence, and Blockchain can help insurance companies improve their fraud detection and prevention capabilities while reducing operational costs.
The writer is a senior manager at PwC Kenya.
