The generative AI revolution in the insurance industry

In the data-driven insurance industry, technology has always been a key driver of progress, from the use of telematics to paperless customer interactions that provide a more efficient, accurate, and tailored service. As the industry continuously adapts to digital transformation, new technologies such as generative AI (GenAI) are paving the way for even greater advances. Based on the findings of PwC’s 27th Annual Global CEO Survey and the challenges and opportunities identified, this article explores the integration of GenAI in the insurance industry.

Strategic integration in insurance

For the insurance sector, GenAI offers numerous opportunities to redefine service delivery to improve risk assessment, claims processing, and customer interactions. Using GenAI, insurers can analyse large amounts of data more efficiently, enabling more precise risk evaluation and tailored policy offerings. In addition, GenAI can automate routine tasks, freeing human agents to focus on complex claims and personalised customer service, increasing both efficiency and satisfaction.

Our use cases

Many of our clients have solutions that are developed and hosted on proprietary software. A good example is the use of MATLAB or Prophet to calculate actuarial reserves. The industry is moving towards hosting these solutions in the cloud, leveraging open source coding languages such as Python and R. Model conversions require specialist knowledge of proprietary software and can be very time-intensive and error-prone. Good code documentation is time-consuming and often an afterthought for such programmes.

GenAI coding assistants, such as GitHub Copilot, can help efficiently convert and refactor code from one programming language to another. It can also help with code optimisation and documentation.

Key benefits of GenAI coding assistants include:

increased speed of system migration;
involvement of a wider range of resources, reducing reliance on scarce resources to perform model conversion.

In today’s dynamic insurance landscape, insurers face a significant hurdle: outdated legacy systems that are ill-suited for the digital age and unable to adapt to modern customer-focused models. This often results in poor quality input data from policy and claims management platforms, requiring extensive manipulation to standardise. From creating model point files to correcting data errors, this process demands careful attention. Data may have different issues at different valuation periods, and any problems can severely impact daily workflows. Moreover, navigating GDPR and DORA regulations adds another layer of complexity, necessitating compliance while maximising the relevance of the data.

However, optimising the use of data offers substantial rewards. It promises to enhance the customer experience by providing deeper insights and personalised services. Furthermore, effective use of data enables more accurate risk assessment and improves capital and liquidity management. Through process intelligence tools such as Appian, GenAI can transform data management, by automating data analysis for faster, more accurate assessments. In addition, GenAI can quickly develop clean files, suggest data fixes, and provide data quality reports.

The key benefits of GenAI-optimised data management are:

reduced human error in risk assessment;
automated data analysis and model development for faster processing.

Governance and ethical considerations

The adoption of GenAI is not without its challenges. It is important to note that in both use cases, GenAI is an assistant, not a replacement for human skills or judgement. Specialist involvement is still required for design, review, adoption, and testing. As the insurance industry embraces this technological advancement, it must also address and mitigate the risks associated with AI, particularly in terms of improper use that can lead to biased results. Ensuring transparency in AI-driven decisions will be paramount, requiring a balance between using AI for efficiency and maintaining both ethical standards and AI governance guidelines.

GenAI is susceptible to hallucination, which means it can generate misleading or false information. Therefore, output reviews by subject matter experts are a fundamental necessity. For example, GenAI may introduce subtle but devastating errors that manifest themselves in the medium term and cannot be easily identified.

If not trained properly, GenAI may make connections that are either unethical or illegal. For example, if GenAI is used in the context of insurance pricing, it could start to discriminate against future policyholders based on less obvious data sources such as surnames, which may imply a customer’s religion or origin.

How to de-risk GenAI challenges?

Use complex neural network models to understand risks, but use simpler, more transparent models such as generalised linear models to ensure control over the models.

Use interpretation tools: Take advantage of available tools such as SHAP (SHapley Additive exPlanations) that assist developers in interpreting the decisions made by generative AI models. These tools may include feature importance plots, attention maps, and other visualisation techniques.

Prioritise referenceable and auditable training data: Ensure that the training data used for generative AI models can be traced back to reliable and trustworthy sources. This enables easy referencing and auditing when needed.

Regulations for responsible implementation

In March 2024, the European Parliament approved the Artificial Intelligence Act (EU AI Act), which ensures safety and compliance with fundamental rights. Once published in the Official Journal of the EU, the law will enter into force on the 20th day following that date (expected by May 2024). It will be fully applicable 24 months after its entry into force. The Swiss Federal Council has a vested interest in harmonising with EU regulations such as the AI Act to facilitate smooth operations. It aims to develop a policy framework by the end of 2024, identifying necessary actions and potential options for sector-specific and, if required, cross-sectoral measures.

To effectively address these challenges, insurance companies should adopt a framework for responsible AI implementation. This includes establishing clear guidelines for AI use, investing in bias detection and mitigation tools, and fostering a culture of ethical AI use within the organisation. By doing so, insurers can harness the benefits of GenAI while maintaining their commitment to fairness, equality, and regulatory requirements.

This stance underscores the strategic balance between fostering innovation and adhering to comprehensive regulatory frameworks. Swiss insurers are poised to navigate the EU AI Act by leveraging their advanced compliance infrastructures and continuing to innovate in ethical AI applications. This approach not only ensures access to the EU market, but also positions Switzerland as a leader in the deployment of ethical AI in insurance.

Summary

The insurance industry has experienced disruptive technological change before. While GenAI can improve data management, automation, and decision-making, it is critical to evaluate its benefits against its limitations and biases. This requires a strategic approach that considers the ethical implications and a commitment to responsible use. For insurance leaders, the way forward is not only to embrace the technological possibilities of GenAI, but also to address its challenges with foresight and integrity.

Contact us

Prafull Sharma

Partner, Cloud & Digital Leader, PwC Switzerland

+41 58 792 18 72

Email

James Norman

Partner Actuarial and Risk Modelling Solutions, PwC Switzerland

+41 58 792 26 13

Email