The Digital Dilemma: Decoding the Challenges in Commercial & Specialty P&C Underwriting Automation
Why is it so hard to digitise the Commercial & Specialty P&C Underwriting process? The world has been on a digitisation journey for the last several decades. The insurance industry has typically been a ‘follower’ industry, observing technological developments and adopting relatively mature technologies as they are proven across other industries.
Even within Financial Services it is typically the banks that lead the various waves of new technologies with insurance following. Nevertheless, the industry has made significant investments in the digitisation of its various processes. Despite these investments the Commercial & Specialty P&C pre-bind Underwriting process is still highly manual and effort intensive utilising a mixture of email, unstructured documents, excel and basic workflow technologies and controls.
According to mea research, carriers spend circa $11bn per year, processing broker submissions for new business, renewal and endorsements. This spend includes activities such as classifying emails from brokers (is it a new business, renewal or endorsement request, is it the first email or a follow up etc.), extracting and saving documents as PDF, word, excel, reading through the content, extracting data items that are needed for the Underwriting process, typing data into carrier systems, performing validation processes such as clearance and sanction checks.
Applying risk appetite rules, performing cat modelling and exposure management activities and prioritising submissions according to other factors such as broker track record on successful conversion of quotes. So why has the Insurance industry not digitised this process yet? The fundamental issue is that the incoming broker submissions are largely unstructured without any standardisation on the format of these emails and documents. Application forms change by line of business, broker, geography and over time (although ACORD standards do bring a level of standardisation in US business).
Loss runs and SOV’s are in excel but the column order, format and content continually changes. Key information needed to Underwrite the risk is often in the body of the email. Risk assessment documents (such as property engineering reports) are voluminous and largely unstructured. On the carrier side, underwriting support operations are fragmented across geography and line of business with the result being lots of small teams supporting the process. Even within each Line of Business, there are often lots of ‘product flavours’ each with different data and Underwriting processing requirements. The largest carriers can have 500 or more of these product flavours.
The industry has tried to solve this problem in various ways with limited success so far. The London Market has launched PPL to enable electronic trading, including the exchange of structured data between brokers and carriers (but for only a small subset of the data fields needed for Underwriting decisions). The initiative on Core Data Record continues the same path by setting minimum structured data requirements for broker submissions, which will help but again unlikely to cover all the fields needed to Underwrite risks across all the different lines of business. Various brokers have created digital trading platforms that carriers can ‘API’ into – but they only cover a small fraction of the business traded globally.
On the carrier side, individual carriers have launched many initiatives over the last several years, to deploy various forms of AI to interpret the content received as part of broker Submissions and extract the information of interest from an Underwriting perspective, enabling auto-population of the various systems involved (e.g. submission logging, clearance, cat modelling, exposure management, rating systems etc) in the Underwriting process thus digitising & largely automating the process of preparing submissions for an Underwriting decisions (which of course will continue being a human decision).
Carriers have used Intelligent Document Processing software providers to help with the task of training their own AI models to be able to deal with all the complexity of receiving broker submissions in thousands of different formats which continually evolve, dealing with the complexities of the same piece of information being expressed in multiple different ways, different data items to be extracted per ‘product flavour’ and then doing that across all the languages that they do business in.
These initiatives have largely stalled – primarily because the time, effort and cost of doing this just does not stack up – the business case does not work. Each field to be extracted from these Submissions needs its own document preparation, AI annotation, AI model training & testing and you have to be able to do this in a way that deals with all the variations mentioned above. The nature of Commercial & Specialty P&C is such that various carriers are receiving the same broker submissions, and each will process those in a different way. Carriers are trying to solve the same problem, which is an industry wide problem, individually. All in their own different way.
This industry challenge is one of the reasons mea was founded – to solve these intractable industry wide problems once and for all. mea’s AI platform is pre-trained across all the major coverages & data items that underpin the vast majority of Commercial & Specialty lines of business on the market today, across multiple languages, thus enabling carriers to deploy such digital processing capabilities in a fraction of the time and cost that was previously possible.
Talk to us if you’d like to know how we managed to deploy mea’s Ingestion module extracting all the fields needed for Submission Setup, across all the major Lines of Business, in several countries of a top 10 Global Commercial & Specialty P&C carrier in under half a year, processing the vast majority of their global submissions volume. mea – making life easier.
