Försäkringskassan

Path to AI Way of Working

Försäkringskassan leads the way in implementing digital solutions for the people of Sweden. Försäkringskassan's strategy to use AI to increase efficiency and legal security has been ongoing since 2017, and in the summer of 2019 Sogeti was chosen as a partner to further their AI program. Sogeti has since had a long relationship with Försäkringskassan and has done several AI projects over the years.

Ensuring Compliance of Sensitive Data

Försäkringskassan handles a massive flow of data, information, and communication - all of which must strictly adhere to regulatory and legal standards. This posed a significant challenge when testing the applications and systems, especially considering the highly sensitive nature of the personal data - ranging from dental records to symptoms of illness. The real customer data could not be used for testing. The conventional data generation methods such as scripts were not good in capturing the unique traits and specific nuances needed for a realistic dataset. It was important for Försäkringskassan to find a solution that would enable efficient testing with accurate data, ensuring quality and compliance in their operations.

Synthetic Data Solution Safeguarding Data Security

Together with Försäkringskassan, Sogeti developed a solution in the form of a scalable microservice powered by advanced machine learning algorithms. This solution could generate tailored texts for various purposes, ensuring compliance with regulations by avoiding the use of sensitive production data. The primary objective was to utilize synthetic text for training AI models, providing an ethical and sound approach, all while upholding the highest standards of innovation and data privacy.

Using a type of model called "Generative Adversarial Network" (GAN), a sample of real data can be fed into the model which is trained to learn the characteristics of the data. The output of the trained model is then a generated synthetic dataset that closely matches the original data in terms of statistical similarity and distribution. The generated data preserves all the properties, correlations and characteristics of the original data so that it can be substituted for actual data. All types of data could eventually be generated – from images and journals to documents and text files without personal data being exposed.

In addition to the synthetic data and its usefulness in testing, the solution can help mask text to anonymize data and create better conditions for developing new services and benefits using AI methods. The synthetic data can be used in training models for text classification, chatbot Q&A and document generation. The solution can also filter, train and generate realistic data. The solution was implemented in the existing production environment and integrated with other smart services.

AI & Innovation Service Center (AISC) Delivering Results

Sogeti mapped the use cases in the innovation maturity model and set up an AI & Innovation Service Center (AISC) team. The AI Way-of-working model allowed for quick installation and implementation. Hackathons were used to gather ideas and develop different techniques.

Försäkringskassan was impressed by the speed of implementation and installation of AISC. Use cases were delivered in different areas of AI:

• Text and image analysis
• Natural Language Processing (NLP)
• Sentiment analysis
• Generation of synthetic data
• GDPR compliance analysis

Adding to the tangible result, the new business understanding of AI performing advanced tasks was an important learning.

Below you will find a short video about the AI innovation at Försäkringskassan. Have a look!

 

 

AI-Enhanced Social Welfare

Försäkringskassan's collaboration with Sogeti exemplifies a remarkable success story in leveraging AI to transform public sector operations. Through innovative solutions like synthetic data generation and GANs, the authority achieved heightened data security, scalability, and compliance. The rapid implementation of AI use cases through AISC showcased not only tangible results but also a deeper understanding of AI's transformative capabilities. This journey reinforces the potential of AI to empower social welfare institutions with efficient, secure, and forward-looking solutions.