Automating Re/Insurance Underwriting

Making re/insurance policy contracts computable is essential to extensibly and scalably automating pre-bind underwriting-analysis as well as post-bind policy-and-portfolio management, explains Anant Borole, Product Strategy Manager at QOMPLX.

Insurance contracts, and more so reinsurance contracts, are a complex, natural language mashup outlining coverages and events that would trigger coverage. Additionally, they contain conditions that affirm or exclude coverage upon the occurrence of an event and over the lifetime of a policy. Contracts are also descriptive of the insured-entity (“insured”). This information and insured description is valuable to portfolio managers and pricing managers.

Pricing managers are tasked with preserving an industry-competitive return on capital. In doing so, they categorize insureds by levels of risk so that an insured’s premium-contribution to a book is commensurate to its risk-profile. For small and medium business (SMB) cyber-insurance pricing, rising cyber-security and technology risks deliver a triple whammy. Growth in premium volumes raises the stakes, the risk-environment alters on an hourly basis and the highest-risk insureds are technology-and-data intensive companies. Such companies have rapidly-iterating “fail fast” business models which, combined with a repeatedly altering risk environment, results in a similarly dynamic risk-profile. Dynamic variance in cyber-security and technology risks is also one of the causes exacerbating recent, substantial market hardening in professional lines, e.g. D&O cover. In light of these challenges, extensibly automating the entire underwriting-analysis chain is key to sustaining a business-process that, accurately and currently, categorizes and prices new quotes.

Portfolio managers are tasked with regulating concentration of exposure, often across several, disparate books of business. Most pressingly, short-tail lines such as property-catastrophe reinsurance require accurate, highly data-intensive modeling and monitoring on a nightly basis. During compressed “renewal seasons”, or as a windstorm approaches, or immediately following a natural or man-made catastrophe, the monitoring required may be more frequent and more granular than usual. Scalably automating such a surge in monitoring is necessary for accuracy and timeliness, and has the added benefit of being able to provide vastly differentiated customer-service.

Upstream from portfolio managers and pricing managers, an underwriter is tasked with incorporating ever-finer portfolio or book-level decisions, from the above processes, into his/her risk-selection process. For him/her, more information is not necessarily a good thing. His/her attention is a limited, highly-valued resource and must be focused on information-processing that benefits from, and aids, his/her expertise and judgment. Automation here helps reduce this information-processing burden by autonomously making and executing transactional-decisions within an underwriter’s workflow.  

On all counts, automation requires a fundamental rethink of how diverse data is captured, organized and analyzed throughout underwriting.  In turn, doing the data-management correctly begins with, and centers on, organizing the data contained within a policy-contract. Data within each contract must be organized, from the outset, such that it is efficaciously amenable to computation-intensive modeling, querying and reporting. This is because modeling, querying and reporting underlie every automatable decision-making task across varying business processes. In short, contracts must be “computable” to be extensibly and scalably automatable.

To this end, the re/insurance and technology practitioners, at QOMPLX, collaborated to build an ecosystem of technologies that support our Contract Definition Language (CDL). CDL facilitates the declarative specification and programmatic access of a portfolio of re/insurance contracts. This powerful approach forms a domain-specific knowledge-base underlying the entire breadth of the underwriting process, and is specifically built to be computable. More specifically, the CDL:

1. Enforces rules for programmatically composing various forms of re/insurance contracts with their underlying policies and coverages
2. Defines a data model to store
   a. insured’s characteristics, policy terms and insured-values for contracts and their underlying
   b. externally-sourced data with utility to a subsequent automatable task, occurring during that contract’s lifetime
3. Defines method-objects, to encapsulate re/insurance domain-specific contract behavior, e.g. calculating deductibles and limits
4. Robustly and efficiently evaluates multi-perspective expected-loss from scenario-explorations/ sensitivity-analyses

Practitioners at QOMPLX have been hard at work, making our contracts computable and, atop that, rebuilding the re/insurance underwriting-analysis process, over the past 4 years. Fortunately for the re/insurance industry we serve, we have had an active ongoing engagement, and strong philosophical alignment, with the aspirations of Oasis LMF, Lighthill Risk Network and the wider London Market. To complement and expand our own efforts, we collaborated extensively with John Cummins (Innovation Partners Ltd) over the course of two years, as he independently investigated the industry’s scope and need for computable contracting. His exercise was funded by Lighthill Risk Network, under the sponsorship of Dickie Whitaker (CEO, Oasis LMF and CEO, Lighthill Risk Network). Earlier this year, Lighthill Risk Network published a report with Mr. Cummins’s findings which also highlights some of our work, in computable contracts, and other contributions to the re/insurance community at large.

Please give it a read and let us know what you think.

<<Get the Lighthill Risk Network Report>>