Case Study

Reducing the Cost of Fraudulent Claims with Aioi Nissay Dowa Europe

Identify, prioritise, and investigate fraudulent claims with continuously learning AI.

650dab230f5bb790d3859f04_ms-and-ad-logo

 

2
%

saved on AND-E's capped indemnity spend in 2022.

800
%

increase in referrals retained by the fraud department.

120
%

increase in the detection of fraudulent claims.

Problem:

Fraud is an industry-wide problem that now accounts for nearly 40% of all crimes. In the UK alone, insurers detected 89,000 dishonest insurance claims in 2021, valued at £1.1 billion, with motor insurance claims being the most common.

Aioi Nissay Dowa Europe (AND-E) previously used a rule-based model to automatically flag and triage a subset of claims for further human investigation. Though this saved many hours, the number of cases sent to triage remained high, the model performance deteriorated over time, and the lack of explainability made it nearly impossible for investigators to budget their time and know what to investigate first.

Solution:

Mind Foundry built a unique fraud detection and prediction solution based on AND-E’s specific requirements that could not be met by any off-the-shelf solutions. The Mind Foundry Platform empowers claims experts to identify, prioritise, and investigate fraudulent activity.

The solution accelerates the investigator’s ability to generate reports and close cases by integrating with external data sources, like Guidewire and ingesting over 7 years of data, including 20 million unstructured documents underpinning fraud ring detection. It automates the prioritisation of each case, using new and previously learnt features, including hand-written notes, phone numbers, registrations, and more.

The most relevant claims get flagged and prioritised in an investigations dashboard, so AND-E can resolve them in order of the likelihood that they contain actual instances of fraud. Other search features uniquely relevant to the fraud case are available in the handler's view, accelerating the ability to close each investigation quicker, accompanied by reports, supporting evidence, and investigator notes, all of which were previously accessed through a disparate set of tools or not accessible at all.

As new patterns emerge, using a Continuous Metalearning capability, the solution automatically and continuously integrates data back into the model to improve performance over time rather than suffer model drift, misalignment, or other modes of failure associated with model deterioration over time. New risks can be effectively governed while enabling the model to learn new types of fraud in production, meaning no manual retraining is needed to continuously adapt, optimise operations, and learn with AI.

Results:

Experts at AND-E have the enhanced ability to investigate the most likely cases of fraud first, as well as interrogate the data behind the Machine Learning predictions, due to the model's explainable and transparent nature. The solution has:

  • Increased referrals retained by the fraud department by 800%, so handlers spent less time on false positive cases.

  • Increased the detection of fraudulent claims by 120% compared to the legacy system.

  • Saved 2% on AND-E’s capped indemnity spend in 2022 and tracking to double that in 2023 with a 4% saving.

All of this plays back to that important piece of keeping premiums down for AND-E’s customers and opening up more opportunities for Machine Learning. The impact of this collaboration extends beyond AND-E and Mind Foundry. The model's features involve data contributions and ingestions from various national fraud databases, such as CIFAS, SIRA, and IFB.

The model's results improve through constant updates from external sources which are then fed back directly into these databases. Insurers across the country use these databases to better inform their fraud identification process and benefit from the improved data quality provided by Mind Foundry's solution.

AND-E Logo

"Mind Foundry’s fraud solution enabled AND-E to improve fraud detection by 120% within the first month of adoption, compared to our traditional system. We expect this to continue improving as the model continuously learns and we expand our partnership. This is hugely beneficial to both AND-E and our customers, helping us to eliminate more fraudsters and become more competitive by reducing the cost of fraud being passed onto our customers."

Nika Lee
Chief Underwriting Officer

Contact us to learn more...

 

Deepen your knowledge

View Resources

5 min read
Why AI Governance Matters in the Fight against Insurance Fraud
Fraud is one of the most pervasive threats in the insurance industry. Insurance fraud is estimated to cost insurers £1.1 billion in the UK alone. As...
4 min read
7 Steps for Scaling AI Governance in Insurance Pricing
The complexity of modern pricing models, coupled with the regulatory changes that are due to come into effect this year, means that insurers need to...
4 min read
Aioi R&D Lab - Oxford in 2024
One year into its mission to solve global-scale societal issues through the creation of products and services powered by AI technologies, the Aioi...

Stay connected

News, announcements, and blogs about AI in high-stakes applications.