Client story: GF FORSIKRING

Identifying 76% of previously undetectable fraudulent claims

Using superior AI models to detect fraudulent household insurance claims

GF-Forsikring (GF) is a Denmark-based insurance company that uses machine learning to optimize its internal and external processes. Aiming to become the hub for machine learning and AI across the island of Funen, GF’s business model relies on AI-driven applications that provide superior customer service and reduce risk. One such model was built by 2021.AI as a proof of concept (or PoC) in 2019.

GF success story

Download the case

Download the PDF-version of “Identifying 76% of previously undetectable fraudulent claims” client story

The challenge

AI applications are increasingly critical across the finance and insurance sectors, which rely on AI-generated models to predict customer success, churn and pricing behaviors, as well as fraud and risk assessments. However, such applications must comply with the Danish Financial Supervisory Authority (FSA) and other regulatory bodies.

In GF Forsikring, with its ambitions for its data science team and AI in general, we see a client operating at the forefront of the insurance industry. We look forward to working with the team as they leverage the Grace platform to deliver production-ready AI models across their business.

Björn Preuß, Lead Data Scientist, 2021.AI

The solution

In 2021, 2021.AI collaborated with GF to create an AI-derived model that could simulate fraud patterns among household insurance claims. Once found successful, the model was used to create a new, second system: an anti-fraud model fueled by machine learning and integrated into our signature Grace AI Platform.

Moreover, we supported GF in establishing the necessary structures and organizational support necessary to run such models. Once familiar with the Grace AI Platform, GF’s data science team then tweaked the fraud model so it would fit GF’s internal systems. Finally, enhanced by Grace’s built-in Governance capacities, the anti-fraud model entered production.

The results

The model accurately identifies 76% of previously undetectable fraudulent claims and can forward suspect claims for review. In addition, the frameworks employed to design the fraud and anti-fraud models remain in use and can be improved as needed, enabling GF to build increasingly complex models and further strengthen its market position in the future.

Project highlights

  • Fraud detection with >76% accuracy
  • GRC support that helps AI models comply with regulations
  • Frameworks allowing GF to pioneer the use of Insurance AI across Denmark

Find out how we can also help you realize success with AI