Länsförsäkringar is leveraging artificial intelligence (AI) to enhance customer service and streamline operations. With nearly half of Sweden's population as customers, maintaining personalised service is vital, prompting the exploration of AI to automate tasks and simplify access to information. By building its own an internal chat client (LF-GPT), similar in function to ChatGPT from OpenAI, they aim to empower employees whilst also ensuring data security of internal information. The implementation of LF-GPT has already shown promising results, improving productivity and raising awareness within the organisation. Looking forward, Länsförsäkringar plans to expand AI applications, particularly in machine learning, to further enhance their services. Despite acknowledging risks, they believe that neglecting AI's potential poses a greater risk. Through careful implementation and adherence to regulations, they aim to harness AI to meet evolving customer and employee expectations.
Länsförsäkringar is a Swedish alliance of 23 regional customer-owned insurance companies. Founded in 1801, with 8,500 employees and serving 3.9 million customers, Länsförsäkringar offers banking and insurance services for private individuals, businesses and farmers. Its local presence provides the organisation with a competitive advantage, which is why it prioritises delivering good, personalised service.
However, with almost half of Sweden's population as customers, ensuring a high level of service to its growing customer base is a resource-intensive task. To free up time to maintain this personalised level of service, processes must be automated and access to information simplified - a task that is well suited to artificial intelligence (AI).
As an industry with processes that typically traversing through large quantities of data and involves numerous administrative tasks, AI has huge potentially to revolutionise the insurance industry. AI technology has become increasingly accessible, offering substantial value through enhancing customer experiences and boosting productivity, quality, speed, and precision in a cost-effective manner.
Starting the journey to build an internal GPT
When the business benefits of AI (and notably, Generative AI) began to be widely recognised in 2023, Länsförsäkringar decided to launch a test project and build a “chat client”, also known as a chatbot. Senior management initiated an objective to explore and build knowledge and awareness around how a Generative Pre-trained Transformer (GPT) and Large Language Models (LLM) could potentially “supercharge the knowledge worker” throughout many parts of the organisation.
Working in close collaboration with Microsoft, Länsförsäkringar built its own GPT chat client, LF-GPT, which all of the company’s employees have access to. The chat client, built on Azure OpenAI Services from Microsoft, is similar in function to ChatGPT from OpenAI and currently offers a chat experience based on the LLMs GPT3.5 Turbo and GPT4. There is no additional functionality or knowledge added, and it is a pure “vanilla GPT” experience. Länsförsäkringar chose to collaborate with Microsoft as it is at the forefront of AI and had everything available for smooth enterprise use.
One of the project goals was that an internal AI initiative could act as an innovation catalyst for the entire company. Knowledge and experience with AI were recognised throughout the company as being important for the future, across all areas of the business. The main driver for creating the chat client was to be able to give employees the power of an advanced GPT, but as an internal service that would maintain the data security of internal information.
LF-GPT was built with a small team of four to five people in total, including one skilled developer, that worked part-time on the initiative and started mid-august 2023. The client was in production by the end of September. No one in the team worked full time on the project and roughly half of the timeline was spent on pulling the team together; getting access to the technology; and finalising risk analysis and sufficient documentation. This meant that only around two to three weeks was required for the actual programming of a front- and back-end solution.
Security was at top of mind for Länsförsäkringar, as was compliance with data protection. The insurer also created clear and concise guidelines for employees on how the tool should be used, along with the limitations of the tools and warnings about not trusting everything that it produces. LF-GPT was also limited to the use of non-confidential information, allowing only public and “internal” information to be used. Despite the limitation of being unable to use personal information of customers, various use cases of LF-GPT began to create value for the organisation.
Published May 2024
Use cases of LF-GPT already bringing business benefits
During the beta phase, the team that worked with LF-GPT sent out a staff survey, looking at how employees used the technology. The survey showed that LF-GPT has helped with a range of business activities: from planning workshops, and giving presentation tips; to finding knowledge, generating code and translating text.
The general response from staff has been very positive. The company has noticed that employees are leveraging the technology to support their work and ease workload burden. This way of working is something that is often highlighted as an advantage of AI. The technology can increase individual productivity, give employees better working conditions, and free up time which staff can use to create more meaningful value for customers.
A large percentage of staff in the organisation had no experience of LLM and GPT-technology prior to the initiative. By rolling out this technology widely, Länsförsäkringar was able to raise the awareness and knowledge about the tool and various use cases. Having a workforce with growing knowledge and understanding of the capabilities of AI technology also means that it able to better investigate and evaluate how other AI-enabled technology can benefit the business. For example, there has been a growing interest in the usage of machine learning (referred to as more “traditional AI” solutions).
For a tech-focused company like Länsförsäkringar, the implementation of new technology is also fundamental to being an attractive employer in the future. For the new generation of employees, being able to offer modern working environments with access to AI-supported productivity tools will not only be a matter of course, but a requirement for any business wanting to be successful in the future. During the project, the development of LF-GPT was shared on Länsförsäkringar's career site and on LinkedIn. Interest in its activities was high and the number of job applications increased dramatically.
Looking forward: Seizing future opportunities, but mitigating associated risks…
Länsförsäkringar is already working on version 2.0 of LF-GPT, which will enhance the current version with features including chat history, smartphone availability, and a number of other features similar to the ones that can be seen in OpenAI’s current ChatGPT.
The company is running pilots to see how AI can be used in more specific scenarios across several focus areas, e.g. helping employees respond to customer queries. There is also proof of concept (POC) work being done to evaluate the readiness for use of this technology by customers.
Several more areas have been identified where machine learning (ML) and AI can provide significant benefits. While the results of GPT and GenAI are not always that accurate or reliable, ML is a proven technology. Areas of interest for Länsförsäkringar in ML include improving claims handling and accuracy; improving precision in market and CRM activities; anomaly detection; and improving, supporting or enhancing the models the organisation relies on (e.g. risk, pricing, and reserves).
However, Länsförsäkringar recognises the importance of addressing potential biases inherent in ML applications. Measures must be implemented to counteract these biases. The organisation’s values maintaining a "human brand" ethos, ensuring customers have the option to engage with a human representative when needed.
Despite the risks associated with AI, Länsförsäkringar believes the greater risk lies in neglecting to explore and understand this technology, potentially failing to meet customer and employee expectations. By adhering to forthcoming AI regulations and carefully considering its implementation, these risks can be mitigated.
About Länsförsäkringar
Länsförsäkringar comprises 23 local and customer-owned regional insurance companies and the jointly owned Länsförsäkringar AB. Customers are provided with a complete offering of banking, insurance and real-estate brokerage services through each regional insurance company, each of which is owned by the insurance customers. The 23 customer-owned regional insurance companies cooperate, thereby combining the ability of a small company to adapt to its customers with the strength of a large company.
Länsförsäkringar is Sweden’s largest non-life insurance group and the eleventh largest life insurance group. It has almost 3.9 million customers and 8,500 employees. The company’s task is to offer total solutions based on different combinations of non-life insurance, accident and medical insurance, life assurance, pension saving plans, fund savings and various banking services.
