by Emeline Carême[1], CodeX industry collaborator
Introduction
The natural language contract has been the cornerstone of insurance processes. It is the product, the common base for discussion, and the materialization of the relationship between the policyholder and the insurer. In the event of claims or a dispute, its terms and conditions can also aggregate both parties’ frustrations. However, investments in the modernization and digitization of the insurance contract have historically been scarce. This has led to scattered contract information and limitations, particularly in insurers’ ability to fully automate underwriting and claims processes and to create new modular and customized products with a short time to market at a lower cost. This was until computable contracts.
Computable contracts – the representation of contract terms in computational form executable by computers – are not a futuristic dream. Rather, they are a current reality – and one that is finding implementation in contemporary commercial practice. In this short paper I will describe a specific use case of their deployment, focusing on the experience of our company, AXA, in developing and deploying computable contracts as part of its business. AXA has created several utilities that allow the creation and application of such contracts, and which have led to demonstrable operating and technical margin improvements through their use. By exploring AXA’s experience, and sharing the reactions of AXA’s senior management, I hope to provide insight from industry on the practical usefulness of the computable contract approach for insurance applications. This paper will also summarize AXA’s efforts to integrate large language model (LLM) approaches with our computable contracts’ approaches. Why computable contracts?
Based on joint research conducted with Stanford University, AXA has been working since 2018 on its own implementation of computable contracts. AXA teams created a logical language to represent the legal wording of insurance contracts. After several years of collecting information on business needs from a number of sources, AXA created a software suite that leverages this representation of the insurance bargain to automate or assist business operations, such as contract underwriting, or claims coverage check.
Today, AXA Computable Contracts (CC) enables the transformation of written contracts (products, policies) into standardized digital models. These models are readable, understandable, and executable by machines. This comprehensive and unified modeling of the insurance cover can be leveraged consistently across IT systems and organizations, driving standardization, interoperability, simplification, automation, and transparency.
A new paradigm across the whole value chain
The solution developed by AXA consists of three main products: Contract Studio, Coverage Check (Instant Coverage Check and Claims Coverage Check), and Contracts Analytics, each addressing a specific segment of the insurance value chain.[2] Each product can be used as a stand-alone proposition or used in combination with the others to unlock additional values.
Contract Studio
The Contract Studio tool seeks to improve productivity of staff for policy underwriting and to give the sales staff a tool to develop live policies together with clients and/or distribution partners. While it previously had taken an experienced operator several days to design and validate a contract in line with customer needs and company policy, this can now be done in a matter of hours. Contract Studio reduces the insurers’ risk by enforcing essential parts of the Underwriting guidelines as the dependencies between clauses, or the use of the latest version of a standard clause. These restrictions are built directly into the tool itself. It can also help to improve the development of new business by allowing the launch of customized products in days instead of months, thus reducing time-to-market and giving the company a competitive edge.
Contract Studio was successfully deployed in AXA’s Partnerships subsidiary in France and has enabled to automate the generation of insurance notices for 15 different B2B2C products. The solution is also in deployment at AXA Switzerland, for policy underwriting in Trade Credit Insurance and will likely be extended to other European subsidiaries in the coming years.
The improvements made possible by Contract Studio have been noted by AXA’s management. As one senior AXA executive [3] has expressed it, “Thanks to the significant time saved per contract, the automation of the underwriting rules, and the future access that will be given to our front sales staff, Contract Studio will allow our teams within our AXA Trade Credit Insurance Community to provide a faster customer service as well as handle a much higher volume of business as we expand into other countries across Europe.”
Coverage Check
Coverage Check provides a pair of utilities. The first product is called Instant Coverage Check. It provides a simulation tool for customers to better understand their coverages pre- or post-sales, for instance to know if they could be covered by the policy for a particular risk, without having to provide a detailed context. It builds customer trust and provides internal benefits such as improved productivity and efficiency of sales agents, faster closings, and opportunity to upsell.
The second product, called Claims Coverage Check, is designed for claim management teams, to automate the coverage check and provide an immediate calculation of the coverage. It enables improved efficiency of claim management, and a reduction of the leakage due to human interpretation of the policy.
As of today, AXA Coverage Check products have been successfully deployed in three AXA subsidiaries:
- AXA Partners automatically executes Claims Coverage Check for two product families:
- Motor & Home assistance products (about 300 product variations) in five countries (Belgium, France, Germany, UK, Italy and Luxembourg) representing about 1,3M claims per year handled by 1000 employees in customer service centers
- Travel products (+3000 active product variations) in three countries (Germany, UK, Spain), handled by 2500 collaborators for more than 300 retail partners
- AXA Switzerland is leveraging Claims Coverage Check for First Notice of Loss and deductible validation in the Invoice check since 2022 representing about 200k claims per year.
- Direct Assurance has deployed the Instant Coverage Check for their Motor product since early 2023 powered by more than 2M calls to our API per year.
AXA’s management has noticed the positive value generated from these tools as well. The CEO of AXA Partners [4] said “We use Computable Contracts at AXA Partners to drive the exchanges with our customers, allowing quicker answer than before and therefore optimizing their overall experience. The tool also enables us to make sure we are not exposing ourselves to leakage, i.e. providing a service out of the contractual engagement. Started few years back at AXA Partners, this significant initiative is a key technical enabler that leverages digital to reduce the number of low-value-added human tasks.”
Contract Analytics
The third product deployed at AXA is Contract Analytics. It performs portfolio risk monitoring, provides market insights, benchmarks insurance products. This automated review is made possible by the data generated by the Computable Contracts approach. The Contract Analytics product provides benefits such as improved risk management (e.g. optimized management of reserves) and improved product performance.[5]
Roland Scharrer, AXA’s Group Chief Emerging Technology & Data Officer, has been enthusiastic about the potential applications of Contract Analytics, declaring “Portfolio analytics is simply a game changer for our industry. Powered by Computable Contracts, it has the potential to unlock significant value by encompassing our sales, exposures and claims history. In an increasingly unpredictable world, this also makes us better prepared and more resilient as an organization as it will revolutionize the way we anticipate the impacts of Black Swan events such as the COVID-19 pandemic.”
Multiple benefits for customer experience and business value
All three solutions have different impacts on the operating margin such as improving efficiency of sales and claims agents, on the technical margin such as reducing leakage and improving risk management, and on the net new business, such as improving time to market, conversion, better customer satisfaction and better performance of products. The following table summarizes the potential benefits it provides:
| CONTRACT STUDIO | COVERAGE CHECK
Instant Coverage Check |
COVERAGE CHECK
Claim – Coverage Check |
CONTRACT ANAYLTICS
|
|
| Impact on operating margin | Improved productivity of staff for product design
Less resources required at constant scope |
Improved efficiency of sales agents Faster closings
|
Improved efficiency of claim mgt agents
Faster treatment of claims
|
|
| Impact on technical margin | Reduced leakage
Less errors in claim management
|
Improved risk management
Optimized management of reserves |
||
| Impact on net new business | Improved time to market
Launch of new products in days vs months |
Improved conversion
Better understanding by the customers / prospects.
|
Lower churn on existing customers / better NPS
Better understanding of value proposition by customers |
Improved market performance of products
Mass customization of products
|
Table 1: Summary of the business value delivered by Computable Contracts products across the insurance value chain
Teams at AXA are increasingly leveraging these applications while also keeping a close eye on the Generative AI revolution underway. There is no doubt that the combination of these technologies contains the capacity to considerably scale current use-cases and create new opportunities.
How LLMs are taking Computable Contracts to the next level
Large Language Models have recently demonstrated impressive capabilities for interacting with customers in natural language, or to assist the understanding of large documents, in particular when combined with Retrieval Augmented Generation technics.
At AXA, we have consequently launched a series of initiatives combining the speed of LLMs to process and organize vast amounts of unstructured data and logic embedded in the contracts with the precision, explainability, automation, compliance, and interoperability of computable contracts.[1]These initiatives include:
- Automating or assisting computable contracts generation from existing Terms and Conditions by leveraging the capabilities of LLM to extract concepts as well as logical rules from natural language; this will be delivered to our customers by the summer of 2024.
- automatically, risk exposure information from policy documents, as an input for risk analytics at policy and portfolio level; this is a Proof of Concept on-going with two AXA subsidiaries.
- Assisting the comparison of two insurance policies among different dimensions, to identify their respective strengths and weaknesses, and capture new customers from competitors; this is an on-going Proof of Concept with one AXA subsidiary.
- Improving the claims experience by leveraging LLMs to extract unstructured data from voice recording, proof documents or form text. This is an on-going Proof of Concept with one AXA subsidiary.
This integration paves the way for more accurate and reliable underwriting risk assessments and claims coverage checks in natural language. This groundbreaking development has the potential to transform the way we approach mass-customization of contracts, with significant impact on product offers, underwriting procedures, claims processes, and risk exposure management.
Conclusion
Computable contracts may seem futuristic, but they have already been implemented in broad areas of commerce. In the insurance industry, AXA has already integrated and used in practice computable contract approaches, witnessing significant benefits from their deployment. The emergence of LLMs complements the utility of computable contracts, rather than replace it. The future digitalization of insurance is likely to be based on a combination of computable contracts and LLMs.
——————————————————————————————————————————————
[1] “AXA: Gen-AI can help to demystify insurance contracts.” Intelligent Insurer, February 8, 2024. Available at https://www.intelligentinsurer.com/axa-gen-ai-can-help-to-demystify-insurance-contracts
[1] Head of Computable Contracts, AXA. The author is grateful for the contributions to the writing of this article made by a number of colleagues at AXA.
[2] AXA. Unlock the Future: AXA Strategic Plan 2024-2026, February 22, 2024, at slide 18. Available at https://www-axa-com.cdn.axa-contento-118412.eu/www-axa-com/a790904f-c607-450c-a76a-eeeb66c1c2b5_AXA_Strategic_Plan_2024_2026_Presentation.pdf
[3] Mariana Mastrogiovanni – Global Trade Credit Insurance Director – AXA Versicherungen AG
[4] Marie-Louise Elhabre, CEO of AXA Partners (AXA’s subsidiary dedicated to designing and distributing insurance & services alongside their business partners’ products)
[5] Ancellin, R. (2022). Automating Risk Analysis in Corporate Insurance Portfolios with Computable Contracts. MIT Computational Law Report. Retrieved from https://law.mit.edu/pub/automatingriskanalysisincorporateinsuranceportfolioswithcomputablecontracts