The use cases for artificial intelligence (AI) in insurance are strong, but the barriers to entry are not insignificant.  

In fact, the Forbes article "Want your company’s AI project to succeed? Don’t hand it to the data scientists, says this CEO" estimates that between 83% and 92% of AI projects fail. Why? Hidden in the corners of inefficient and ineffective systems is poor-quality data full of nonsensical information and inconsistencies. For example, inconsistencies may take the form of one claim on two policies, or a single policy attached to multiple agreements. Poor-quality data such as this can obscure the insight sought through AI integrations.  

Simply put, good-quality data is data that is accurate and complete, e.g., when an insured’s information, coverage type, policy term, premium, deductible, coverage limits, and endorsements are clear and up to date in an insurer’s system. Poor-quality data, on the other hand, is missing values and contains incorrect information or inconsistencies. The consequences of using poor-quality data to train or operate AI models include:  

• Underpricing or overpricing of policies due to incorrect risk assessments  

• Inefficient and ineffective fraud detection  

• Delays, mistakes, and inefficiencies in claims processing  

• Incorrect assumptions about customers, leading to inappropriate marketing strategies and customer interactions  

• Non-compliance with regulations and legal requirements, resulting in the potential for fines, legal actions, and reputational damage  

• Reduced trust in a company’s services  

GETTING ORGANIZED: STRUCTURED VS. UNSTRUCTURED DATA  

Understanding the types of data you have is a crucial step when preparing to incorporate AI into your business processes. This process involves data discovery, assessment, and categorization, and it lays the foundation for building effective AI models and systems. There are two main categories of data: structured and unstructured.  

Structured data refers to information that is organized and stored in a predefined format, making it easy to search, analyze, and process. This type of data is typically stored in databases and can be readily categorized, sorted, and used to gain insights, identify trends, and make data-driven decisions more easily. Structured data is able and ready to be manipulated by AI.  

For example, an insurer’s structured data might include:  

• Policyholder information: Name, address, contact details, policy type, coverage details, premium payments, etc.  

• Claims data: Details about past claims, such as claim amount, date of loss, cause of loss, settlement amount, etc.  

• Underwriting data: Information used to assess risk and determine policy eligibility, such as medical history, driving records, credit scores, etc.  

Unstructured data refers to information that doesn’t have a predefined structure and is not organized in a way that is easily searched or analyzed by traditional methods. This type of data is often in the form of texts, images, audio, video, social media posts, and other formats.  

Being able to review your data and determine what is structured and what is unstructured is the first step in understanding what it takes to integrate AI into your core system.  

3 TIPS FOR MAINTAINING STRONG DATA BEST PRACTICES  

Clean data is key to a successful AI implementation and data transformation. And yet it’s human nature to take shortcuts when under pressure, and the consequences of inserting data midstream can set an organization back in its digital transformation and AI implementation journey.  

Here are three practices that insurers can engage in to maintain strong data practices:  

1. Establish clear data governance policies and procedures to ensure that data is managed consistently across the organization. Include defined roles and responsibilities for data ownership, quality monitoring, and maintenance. These policies and procedures also should extend to third-party partners that filter in data through integration.  

2. Provide training to employees who interact with the company’s core system(s) to ensure they understand the importance of data accuracy and the proper procedures for data entry and maintenance. Help employees understand how their initial time upfront can save them and customers time in the long run.  

3. Conduct regular data audits to identify inconsistencies, inaccuracies, or outdated information. In addition, develop protocols to rectify the issues identified during these audits.  

TAKE ACTION  

Internal and external resources and partners with knowledge of an insurer’s systems and data are critical to the implementation and ongoing use of AI. Partnering with different AI and data management providers can help bridge the gap between what’s possible today and how insurers build on the successful application of AI tomorrow.  

For further information on how insurers can clean up their organizations’ data to make way for AI, read our “Get Your House In Order For…AI” e-book.