22 April 2024 2 min read

🤖 AI for reinsurance contracts: semantic search engine and clause identification use cases

Reinsurance Tutorials #16 - Season 3

Hi everybody 👋


Today, and for the 16th Reinsurance Tutorials video of the season, we will talk about the " AI for reinsurance contracts: semantic search engine and clause identification use cases"


This subject will be addressed by CCR Re experts Madeline Jauvat and Akli Kais.


Let’s start! 

[Akli Kais] : Hello and welcome to this video on AI in reinsurance.


[Madeline Jauvat] : In this tutorial, we will be discussing how semantic search and clause identification engines are transforming the reinsurance industry.


[Akli] :With the arrival of artificial intelligence, the reinsurance industry is undergoing a significant transformation. The traditional methods of managing and analyzing data are being replaced by sophisticated AI algorithms, helping reinsurers to make better decisions and reduce their risk exposure.


Today, we will delve deeper into the two use cases by exploring how they were developed at CCR Re and the benefits they offer to the users. We will also talk about some of the challenges that come with implementing AI in reinsurance and how the digital, legal, and underwriting teams are working together to overcome them.


[Madeline] : In reinsurance, each year, underwriters and legal advisors deal with thousands of diverse contractual documents and often lack time to review the overall portfolio. It is also not uncommon to discover new provisions, policies and clause changes when analyzing their wordings.


In this regard, it seemed essential to collaborate with the digital factory team to create a tool that could automatically analyze the different aspects of a contract, going from identifying and extracting key clauses within a document, and compare them with standard wording determined by the legal advisors, or even search for other examples of redaction that shares the same semantic wordings for a given clause. Moreover, we also wanted the possibility to know quickly whether an ‘abnormal’ clause existed in the portfolio and finally compare different versions of the same article across all our contracts.


These features help us decide whether the document is acceptable for the underwriter to sign or not. If not, the tool will alert the legal team who will deeply dive into the contract and review the identified anomalies.


[Akli] : Developing these tools requires a close collaboration between, in the first hand, the digital factory team, with its ability in developing software solutions, and the business team, in this case legal and underwriting teams, for the business understanding and tool requirements. The development life cycle of these tools was splitted into multiples iteratives steps:


  1. Identify requirements: The first step consists of identifying the requirements of the legal team. This would involve understanding the specific clauses that they need to extract from reinsurance contracts and define the needs for the semantic search engine, such as the types of data, search criteria needed for the filters to be effective, as well as any specific features they require.

  2. Gather data: The next step was dedicated to gathering the data that engines will use. This includes historical contract data, clause wordings with the associated labels. Many iterations are mandatory to ensure that the data is relevant and accurate.

  3. Develop algorithms: Then, the AI Lab implemented the natural language processing and machine learning techniques to understand the meaning behind words and phrases. Fine-tuned pretrained language models were the main techniques used for this purpose.

  4. Test and refine: Once the initial algorithms have been developed, the digital and legal team work together to test and refine the engines. This involves running them on a small set of data and fine-tuning the algorithms to ensure that they are accurate and efficient.

  5. Scale up: Once the clause and search engine has been successfully tested and refined, we ensure that we can scale up to handle larger amounts of data and work on integration with other systems and processes within the CCR Re workflow and ensure that they are fully integrated and monitored through time.


[Madeline] : At the end, such tools present many advantages:

  • The clause analysis provides underwriters with a quick and efficient preliminary legal analysis and alerting, where the semantic search engine helps them find other examples of clauses for their contracts.
  • These 2 tools help legal advisors challenge CCR RE standard clauses and have a global overview of the risk exposure.

[Akli] : Now you know how semantic search ad clause identification engines are transforming the reinsurance industry


[Madeline] : Thanks a lot for watching the video, and goodbye.





Bye for now 👋

 See all videos


📺 More episodes are coming... Subscribe now to receive them by email! 📥