Case Study
IFFCO-Tokio Saves Over $1M Annually by using Fraud detection with Ai
In the insurance business, one of the biggest challenges is dealing with fraudulent claims, which may be difficult to detect, resulting in financial loss. IFFCO-Tokio General Insurance Company Limited decided to turn to AI for help in detecting potential fraud cases involving motor and health claims. With solutions and models built using the Ai Geeks AI Cloud platform, it has managed to sieve out fraudulent claims and expects to save almost 70 million Indian Rupees annually (US$1 million).
Incorporated in 2000, IFFCO-Tokio is a joint venture between Indian Farmers Fertilizer Co-operative Limited, the world’s biggest co-operative reaching more than 50 million Indian farmers, and Tokio Marine Group, the largest property and casualty insurance group in Japan. It offers a wide range of insurance products to 10 million customers across India
The Challenge
As one of the biggest motor insurance companies in the world’s second most populous country, IFFCO-Tokio receives more than 2,000 motor claims every day that are processed across 12 claim service centers by around 100 claim officers, within specified timelines. The process is not only time-consuming, but also demanding. As a result, claim officers often do not have the time to devote much attention to every case to check against possible fraud.
Motor insurance fraud can range from providing misleading details and inflating repair costs, to staging accidents and claiming for fictitious injuries or losses.
IFFCO-Tokio wanted to tap the power of AI to analyse incoming claims and spot anomalies to reduce fraudulent claims that could hurt its bottom line. That same solution can then be adapted to other products, such as health insurance, which face similar challenges.
The Solution
The claim fraud prediction solution that it envisaged must be able to detect fraudulent activity, suspicious links and behavioral patterns based on historical claims data. “We were delighted when we came across Ai Geeks whose Kaggle Grandmaster guided my team remotely on using Ai Geeks’s automated machine learning with pre-built algorithms,” said Seema Gaur, Chief Information Officer and Head of IT of IFFCO-Tokio. “We could deploy it on-premise and use it with a small team, which is ideal for our small team of 3 developers who have only a couple of years of experience. Deploying it on premise is vital for our business due to data privacy,” she added.
Using the Ai Geeks platform, the model gives each claim a score of between zero and one, with one being a higher probability of fraud. With the system flagging such cases instantly and with greater accuracy, the team can pay greater attention to claims with higher scores. Following the successful deployment of the fraud prediction model for motor insurance claims in July 2021, the company then created another model to analyze, detect and predict health insurance claims.
Unlike motor insurance claims, which are done by the policyholder, health insurance claims in India involve a third-party administrator acting as an intermediary between the hospital, which is claiming for services provided, and the insurance company. In most cases, payment is made directly to hospitals.
From the moment a patient is admitted to the point of discharge, there are many triggering points for fraud ranging from medically unnecessary services to overcharging for over-prescription of medicine. A combination of rules and data from a third party is used to train the model to detect such frauds.
Benefits
IFFCO-Tokio is reaping benefits from using the AI models. Based on more than 100 confirmed fraudulent claims detected each month so far, the projected saving from fake motor and health insurance claims are around US$1 million annually
Additionally, the insurance company’s claims officers can now focus on following up on possible fraud cases flagged by the system instead of analyzing every single claim, resulting in faster processing time for genuine claims
In another development, IFFCO-Tokio is also training an AI model developed using Ai Geeks to determine the propensity for customer retention or the likelihood of customers to renew policies.
Since the start of the trial, involving their online customer base, in July 2021, customer retention has increased by three percent.