Improve the Underwriting Process

Underwriting is a critical business because it involves assessing potential risks that the insured is exposed to. It is why underwriters determine the extent of the coverage and the price the consumer is entitled to. They also decide whether to approve the insurance policy. While adding a new policy to their ledger, an underwriter takes on the risk. Because anything can happen in life, there is a lot of uncertainty around whether a claim gets filed or not. Many factors influence the risk of an individual. Some factors may be proxies for less quantifiable information, while others may be things that an underwriter never thought to consider. If underwriters have enhanced knowledge about who they are insuring, they are better positioned to assess the risk involved in adding a policy to their book. It is why effective underwriting must accommodate how risk exposures can change in light of quantitative analysis and forward-looking decisions. Insurers can use technology to improve their own results and decision-making. New technology is causing a cultural shift to the traditional underwriting approach. Agents are abandoning decision-making based on their gut feelings. Risk categories now need to be nuanced instead of just a yes or no.  In order to be profitable and competitive, insurers are looking to increase rigor in underwriting, using a wide variety of data elements from internal and external sources, geospatial intelligence, predictive analytics, and more.

A machine learning model can help give a more holistic view of the insured through company data and third-party sources. A data science model can better classify and determine quantifiable risk factors and help the underwriter manage the complexity. Through a classification model, one can easily investigate if an individual has similar characteristics to others in a certain population. A match with this pattern would enable the underwriter to increase the premium. Incorporating intelligent machine learning models in the underwriting process not only helps train them to view risk factors more confidently but it helps standardized the risk scoring process across a company. Training can only go so far in keeping risk scoring consistent, so machine learning can help prevent insurance companies from taking on too much risk.

Technology is helping underwriters gain more prominent roles in insurance companies. An EY study says that today’s underwriters are more likely to serve as sales executives, decision scientists, customer advocates, and innovators.  IBM conducted a study on how underwriting was transforming with the vast amounts of data and concluded that insurance companies are using machine learning to increase their bottom lines and gain a competitive advantage. They are also reducing expenses and improving efficiencies. They are optimizing all areas of their business from underwriting to marketing to make data-driven decisions that lead to increased profitability. As part of its research, Deloitte found out that over the last year, most insurers understand that investing in these solutions is the key to survive in a fast-changing environment and have invested in data analytics and machine learning solutions. Genpact, in a study, estimates this approach could boost the new business book by 10-20% by increasing bind rates and lower new business loss costs by 1-8% by directing underwriters away from less profitable risks.


Insurance companies are challenged with various issues that hold them from realizing the true potential of data analytics solutions. Deloitte’s recent study among 68 EMEA insurance companies showed that 90% of them struggle to see a positive business case on merely data analytics solutions. Improving underwriting can be very daunting since commercial group insurances are multi-faceted, intermediated, and often more qualitative than personal lines. The related operating models can diversify based on industry, region, client size, and product. Thus, policy wording and exclusions might seem straightforward until they are challenged by litigation and subject to interpretation. Further, achieving and documenting improved results in underwriting performance can take up to several years.

Underwriters navigate a sea of data, software, and file formats to reach a premium for each policy. It’s daunting and comes with many uncertainties. Underwriters have access to a massive data pool that can be used during the underwriting process and can include address records, location information, geospatial data, which are often inconsistent, complex, one-off, independent, and rarely identical even across internal systems, causing a key concern.

Underwriters deal with risk and spend their day doing research, data entry, pricing risks, and ultimately negotiating that premium value.  This may still be inadequate, given the volume and existing processes involved, to translate directly into how much can be ultimately charged for a given premium.  Underwriters need a better way to optimize their work for maximum profitability. Factors that influence a decision include the probability of bind, estimated premium size, estimated profitability, underwriting effort renewal probability, etc. Many insurers are weighed down by these drivers being unaccounted for in the triage process, leading to gaps in some of the best opportunities, ineffective distribution and marketing, and underwriters spending their time on less than optimal activities.

Triaging is another process that weighs down many insurers.  It is often a manual, unsystematic, time-consuming, inconsistent process that relies heavily on first-in, first-out responsiveness to brokers that can suffer familiarity bias and underwriter judgment as underwriters draw on their own experience and the experience of their teammates to predict if an insured will have a claim. Bias is inadvertently involved since every underwriter has a different perspective. Experienced resources may have to spend a daunting amount of time on simple file preparation tasks, leading to misquoting or declination. This process can often be challenging to align with business processes.

Solution Approach

Some major enablers towards large-scale adoption of AI-ML practices and precision CV-based medicine, available through, include dynamic availability of numerous analytics algorithms, models, and methods in a pull-down type of menu, easy management of important issues like data ownership, governance and standards, continuous data acquisition and data cleansing. platform provides out-of-the-box development frameworks.  The project was started with the relevant environments, which were then created automatically. Development images configured based on pre-defined templates were installed on-premises or in a development VM within the infrastructure. This enabled authentication using LDAP, seamless project setup using Bitbucket, Jenkins, and Docker (ensuring build and deployment without software compatibility issues). 

The platform made available by leverages the latest ML and DL tools while preparing models. It includes Pachyderm-based data versioning, Kubernetes, Kubeflow, and Spark-based ML and DL. It also includes an Istio-based service mesh-enabled microservice architecture, and ELK-based monitoring capability, contributing to a reduction in latency time.

The platform can potentially be fed with input data in the form of insurance databases that include size, sales channels, product mix or geography of an insurer (often viewed as cost drivers) and customer data including available social media exposure, details of actual medical conditions reflective of the customer’s actual experience instead of a general overview. We identified input data points such as claims data, enrollment data, prescription data, and member data in our drive to improve pricing and customer service for our insurance client’s group insurance customers and create effective models.’s AI/MLOps platform allows establishing high-end Alluxio and Presto-based efficient data connectivity and collecting data from diverse sources. The details collected were added as exploratory variables by using libraries and analyzed. The attributes obtained were used for categorization (employing Pachyderm-based data versioning) and then performing univariate, bi-variate, and Bag of Words analysis — for both structured and unstructured datasets through xpresso Exploratory Data Analysis (Data and Statistical Analysis).  Different datasets and their different versions were easily controlled and stored into an xpresso Data Model (XDM)-enabled data store that enabled easy retrieval and storage of datasets/ files into internal XDM. This was achieved using two excellent features of

  1. Data Connectivity Marketplace libraries
  2. Data Versioning

We also interacted with actuarial, underwriting, and product teams to discover and understand their nature of costs and current pricing methodology. This helped us develop a predictive model for costs on a cohort basis based on various input data parameters. read factors from a varied recommendation text connection and generated a result from all these variables obtained, models were created and versioned — enabled by MLOps platform. It resulted in a pricing methodology that incorporated the cost prediction model into the current underwriting process. We were able to help insurers identify potentially high claimants, high-risk customers and also save underwriters time in gathering data towards different conclusions while underwriting. They were able to focus more on analysis and decision-making, increase the granularity of risk analysis and enabling pricing adequately, improve underwriting productivity and throughput, achieve greater consistency in decision-making and strong governance of the underwriting process, and reach 82% accuracy in cost predictions. It also provided an opportunity for the insurer to extend enhanced customer service by offering advice and protection from health adversities.

How can help Finance Organizations transform their journey to cognitive AI solutions is an AI/ML Application Lifecycle Management Platform. enables complete lifecycle management of AI/ML solutions, addressing the AI transformation journey of enterprises on any cloud platform of choice. offers functionality essential for building AI/ML solutions – primarily enabling data scientists to rapidly build predictive and prescriptive models. The platform provides a user-friendly interface to develop, deploy, and manage AI/ML solutions at scale. In addition, supports the incorporation of these solutions into business processes, surrounding infrastructure, products and applications.

Key benefits of include:

  • Empowers data scientists to transform AI/ML research into solutions 
  • Improves the productivity of data scientists by enabling them to focus on the business problem, developing algorithms and rapid experimentation of models 
  • Addresses the shortage of skilled data science resources with automated workflows, toolkits and frameworks 
  • Manages AI transformation journey costs without any wastage of R&D efforts  
  • Provides an enterprise-ready and secure environment for complete lifecycle management of AI/ML applications
  • Enables at-scale deployment of enterprise AI/ML applications on-premise, cloud (AWS, GCP, Azure), or hybrid environments

Additional details on can be found at: We can schedule a demo of the platform for anyone interested in learning more.

Have Any Questions?

Need more information about the platform?