How to Achieve Data Readiness for AI: Practical Advice for Insurance Companies from LIMRA, Microsoft and UCT Experts

If you were to start an insurance company from scratch today, I guarantee you AI is going to be embedded across your value chain. It is probably going to be a central element in a lot of your processes.

1. Use cases for AI in life insurance

The insurance industry is on the cusp of a major transformation driven by Artificial Intelligence (AI). Already many insurers are using AI to power important processes that dramatically impact efficiency, customer engagement and decision-making. Common, powerful use cases include:

GenAI Enhanced Customer Experience: Insurers use AI to get more personalized understanding of their customers in an effort to understand how to make their experiences better. AI can customize interactions, automate tasks, and improve risk assessment. These outcomes obviously lead to a more efficient and positive customer experience.

Improved Decision-Making: AI can analyze vast amounts of data to identify patterns and trends, enabling better underwriting, claims processing, and fraud detection. Insurers are also looking at their insurance products and asking, how do you use analytics to make products more connected and digital?

Increased Efficiency: AI can automate repetitive tasks, freeing up human employees to focus on more complex activities and make the employee experience better.

Gen AI technology is relatively new so the deployments that we're seeing at our customers right now are centered on low hanging fruit. They are internal-focused use cases, and many are built around using business applications to make the customer (indirectly) and employee (directly) experience better. “One of the most common questions at the moment is, how can I create a Chat GPT-like experience which allows employees to enter a prompt and immediately get a useful and usable response, and is architected in an environment that is secure and that leverages our data?

2. The ‘born analog’ challenge: legacy system data siloes

While AI offers exciting opportunities, insurers face challenges in realizing its potential because of the prevalence of legacy systems and variable quality of data management practices.

Most insurers are mature enterprises with a history stretching back decades, if not even longer. They are definitely not ‘born digital’. Which means that they are not starting from a clean sheet when it comes to data.

Life insurance and annuities companies globally are on a “becoming digital” journey where hybrid (legacy analog & 1st gen digital) data collection/transformation methods are meeting with purely digital storage and usage models. This does create an interesting complexity while these 2 methods exist within an insurer.

That creates a challenge for those who were born analog.

Analytics, AI, LLMs and other new data-driven applications require consistent and high levels of data quality. And, unfortunately, that’s not the situation most companies find themselves in. They don’t have perfectly formatted data, that’s universally available in real time. More likely, data structure, collection and accessibility will have evolved over time. Some companies refer to it as ‘the swamp’ because there’s probably a mixed bag of formats and structures that have evolved over the years. And a lot, if not most, of it is going to be trapped in legacy PAS and CRM and ERP systems that have been added to the IT environment over time.

This means that in order to build a foundation for future digital applications companies have to find a way to modernize their data and make it ready for new uses. Insurers have to have a data modernization plan for how they will access, cleanse, transform and migrate all the legacy data that’ve accumulated over the decades.

Traditional insurance companies often struggle with:

Data Quality: Legacy systems often contain inconsistent and poorly formatted data, hindering AI applications that rely on clean and organized data. Data Modernization: Modernizing data infrastructure and governance practices that are crucial to becoming more “data-ready” for AI integration. Data Silos: Many insurers have multiple legacy policy admin systems that make it difficult to access data in real time, effectively imprisoning a lot of a company’s data 3. The AI readiness roadmap for insurers In the insurance industry, preparing data for Business Intelligence (BI) and Artificial Intelligence (AI) projects is a strategic imperative. Much of the data can have crossover dual-use with minimal change, but in the AI realm unique considerations have to be applied (sometimes to the same data that is collected for BI uses.) To ensure a seamless transition towards data-driven decision-making, insurers must undertake specific measures:

Vision and Strategy: Tie AI as an enablement technology clearly to your business objectives and business desired business outcomes. Have a time-boxed, realistic plan and execute that plan. Then everything flows from that: your people, your processes. And remember, AI may be a highly advanced technology but, at the end of the day, it is still a tool.

It's about data integration—across systems and with the goal of achieving business objectives. Don't create data sets for ad hoc reasons in disparate formats, isolated in different siloes.

Don't lead with the technology. Lead with the business strategy and business problems that you're trying to solve for.

Be willing to engage in “strategery”! Shifts will occur due to your own personal and organizational learning. We have to be willing to course correct as new findings and experiences drive our fundamental understanding of our organizations need for and use of these powerful new use cases.

Regulation and prioritization

Regulation affects strategy in terms of how insurers need to think about how they prioritize AI projects.

Some of the things that I'm doing with my customers right now is helping them understand where they should prioritize. We're a very regulated industry. That's not going to change. AI is not intended to make decisions. Not right now. What it's intended to do is augment, to make things a little bit easier for people. As insurers prioritize, use cases where regulation and privacy considerations are considerable, won't be safe places to start. A great example is a generative AI-enabled customer facing chat bot, which relies on customer data to elevate experience. We don’t see many customers going here yet.  Whereas if it's internal use case, such as improving contact center representative, claims handler, or underwriter productivity, you're probably going to want to start there. This is where we are seeing most of our customers deploying right now; helping their employees be more productive and effective at work.

4. Data is not just an IT project—Business unit collaboration is critical A robust collaboration between Information Technology (IT) and business is essential for developing a holistic understanding of data needs, which are the cornerstone of effective AI deployment. The integration of insights from both sides of the enterprise ensures that data initiatives will not be only technically viable, but also closely aligned with overarching business goals, thereby enhancing the relevance and impact of AI projects.

Data is not just an IT project—Business unit collaboration is critical A robust collaboration between Information Technology (IT) and business is essential for developing a holistic understanding of data needs, which are the cornerstone of effective AI deployment. The integration of insights from both sides of the enterprise ensures that data initiatives will not be only technically viable, but also closely aligned with overarching business goals, thereby enhancing the relevance and impact of AI projects.

Agility in data initiatives These initiatives also have to be flexible enough to adapt to evolving customer expectations and regulatory landscapes. Continuous communication and refining of data strategies will be required to swiftly respond to emerging trends and challenges. A business and technical understanding of proven solutions for the hybrid co-existence of an insurer’s systems topology, will enable more agile data initiatives and stronger governance patterns in a rapidly expanding future state of an insurer’s data. A collaborative approach ensures that data initiatives remain aligned with business objectives, making AI projects more effective and outcome-oriented.

In the same vein, investing in scalable and flexible cloud-based infrastructure enhances your agility in the face of changing market conditions as AI capabilities evolve. The exponential growth in data volumes, driven by digital interactions and the need for real-time analytics, demands infrastructure that can not only accommodate this growth but also do so in a cost-effective and agile manner.

Scalability of data initiatives Cloud-based solutions offer the scalability and flexibility required to manage these increasing data volumes, providing a foundation that supports the agility needed for AI initiatives. Such infrastructure facilitates the seamless scaling of data storage and processing capabilities, ensuring that insurers can leverage the vast amounts of data required for sophisticated AI applications without being hindered by physical hardware limitations. While scaling is significantly easier than it has ever been (in the world of cloud tech), these decisions must be grounded in purpose and value to avoid “over-spending” and eroding thin budgets/margins.

Data governance Establishing a comprehensive Data Governance Framework creates a backbone for ensuring data integrity, context, and security, which are critical to the successful deployment and operation of AI technologies. A robust data governance framework lays down the foundation for managing the vast and complex datasets that AI systems rely on, ensuring that data across the organization is accurate, consistently defined, and securely managed. Coming up with structured ways to deal with and govern unstructured data is a new challenge that many insurers have or will face in their future governance considerations.

Data management roles Central to this framework is the clear definition of roles and responsibilities related to data management within the organization. A clearly defined data organization chart ensures that your overall data strategy will be managed at a high enough level in the organization to give appropriate oversight, while ensuring that, throughout teams and projects, every aspect of the data's lifecycle, from creation and storage to access and deletion, is properly managed.

Data access and ownership policies The governance framework should also detail data access controls and data ownership policies. These measures safeguard sensitive information from unauthorized access, ensuring that data is accessed on a need-to-know basis and used responsibly. Access controls and ownership policies also help in delineating the boundaries of data usage, ensuring that regulatory compliance is maintained, and that ethical guidelines are adhered to.

A well-structured Data Governance Framework protects the organization's data against internal and external threats, creates a single source of truth, and a solid foundation for leveraging AI technologies effectively. The framework enables insurers to harness the power of AI while maintaining the highest standards of data integrity and security.

The importance of data literacy as a foundational capability Data literacy is, at its core, the ability to tell good data from bad data. And data quality is predicated entirely on the human beings that manage that data. Which means data literacy has to come before AI literacy—and that’s across the organization.

This is now considered a core business competency for any organization that conducts business in a digital capacity. If you don’t make an appropriate investment in data literacy, you’re going to have increasingly poor data quality because the people who are on the front lines with the data all day, every day, cannot understand or interpret the context within which the data is being used. And the result is that the bad data is going to permeate and infect your AI models.

So, as a general principle focus on data quality, but also prioritize improving data literacy through continuous learning programs to equip your employees with the skills they require to discern good data from bad data. Invest in AI literacy initiatives so they understand how to work with Machine learning programs and what impact the decisions they make with the data they collect, identify or use to train these algorithms can have.

As an example of data literacy impacting AI literacy, I was talking to an analyst and he started implementing intelligent document processing into an organization. Once it went into production people had to start manually categorizing and classifying the different documents. And he was shocked because on the first day they saw that about 40% of all the documents were incorrectly classified and they didn't understand why.

5. The process for preparing data for AI

Most insurance companies have already begun their data modernization journey. They may be extracting data from legacy sources to put into any new storage mechanism, to cleanse it and modernize it and do things with it to improve your client and customer experience or generate reports. The pitfall is to apply a “one size fits all” approach for your AI data approach/strategy. Insurers have to look at their data differently than ever before to yield amazing results from this amazing shift in technology.

Data assessment and modernization Preparing data for AI applications begins with extensive data modernization. In this step the goal is to assess data quality, remediate inconsistencies, gaps or errors all while standardizing the formats and structures (where appropriate). It may include augmenting data from work around processes and translating code values to "real business language".

Harmonizing the way data is represented and contextualizing the selection/storage enables more efficient data integration and processing by AI algorithms. This uniformity is crucial for AI systems, which can require consistent and structured datasets to learn effectively and deliver meaningful outputs.

Data aggregation Data aggregation techniques collect information from various sources into a unified dataset. For AI, this consolidated view is invaluable, providing a comprehensive dataset that reflects a wide spectrum of variables and scenarios for the AI to analyze. This not only improves the quality of the insights generated but also enhances the AI model's ability to generalize its learning to new, unseen data.

Data migration

Data migration, in the context of preparing for AI use, involves the strategic movement and consolidation of data from multiple sources into a centralized repository, like a modern policy administration system able to feed sophisticated AI algorithms.

This process is crucial for ensuring that AI models have access to a comprehensive and diverse dataset, enhancing their ability to learn, identify patterns, and make accurate predictions.

Migrating data effectively requires meticulous planning to ensure data integrity and consistency, which includes mapping data from various formats and structures into a unified format that AI systems can easily process. Moreover, data migration project are also the perfect opportunity to strategize around how to best cleanse and normalize data to remove duplicates and inconsistencies, which could otherwise skew AI analyses and outcomes. The goal is to create a rich, clean, and structured dataset that enables AI models to operate efficiently and effectively, ultimately driving more informed decision-making and innovation.

Through careful data migration, organizations can leverage their data assets to their full potential, paving the way for AI-driven transformation and competitive advantage.

Data integration

Data integration is an important solution to the problem of accessing data tapped in legacy silos for carriers who don’t wish to conduct data migrations or haven’t yet achieved “PAS nirvana” of a single policy admin

Integration involves using tools to extract & transform data from legacy platforms as it is needed so that it can be shared with apps or solutions that need real time access to the information without as much overhead of the “load” portion of the ETL. As an example, an agent and client self-service portal will need to be to access all client policies on demand regardless of whether they are stored on a modern PAS with APIs connections or are siloed on old mainframes. Data integration enables those portals to call for policy data from any legacy system for review by an agent of client. And the integration can be ‘bi-directional’ in the sense that if a client wanted to change their address for instance, they can make the change on the portal and the updated data can be returned to the original PAS in its new form.

In this way, data integrations give carriers total access to their key data, helps provide 360-degree views of their customers—and does it all in real time, without the need for large enterprise solutions and their associated potential risks and costs (one-time setup/ongoing).

Integrations have a number of high-value uses for life insurers. They enable the rapid deployment of advanced analytics and AI, which can transform the raw data into insights that improve decision-making.

They also make it possible to connect legacy data sources to modular new applications that can be built on a microservices architecture. This means insurers can build ecosystems of smaller, connected apps that perform specific functions, connected by APIs and driven by real-time data unlocked from their legacy siloes. These apps are scalable and can be built to enhance carrier capabilities in every area, from underwriting to claims.

Metadata

Implementing robust metadata management is the final piece of the puzzle. In the AI domain, metadata not only provides context about the data but also about the data's lineage, quality, and the parameters used in AI models. Effective metadata management supports greater data transparency and facilitates collaboration among data scientists, developers, and business analysts. It ensures that all stakeholders have a clear understanding of the data's origins, characteristics, and constraints, which is critical for developing, deploying, and maintaining AI models successfully.

Utilizing these data modernization practices makes it possible for insurers to access the right data at the right time and ensures that it is available for emerging analytics and AI use cases.

Wrap up

Engage in pursuing AI because in 3-to-5 years it is going to be mainstream. That means you need to be thinking and working today on how to manage the changes it will bring to your organization. “AI capabilities and the new abilities employees will have when their roles are AI enhanced are going to require new ways of thinking, new kinds of people, new kinds of skills in your organization. So be ready to evolve and adapt very quickly.

Life insurers are well aware of and hungry to take advantage of the potential benefits offered by advanced analytics and AI, but they face very real challenges in readying their data for new use cases. They continue to struggle with legacy systems that silo historic policy data and the poor quality of, not just a lot of their data, but also their data management practices.

By systematically addressing these steps, insurers can lay a solid foundation for BI and AI projects, ensuring that data becomes a strategic asset driving informed decision-making within the organization.

Some content and quotes for this article were sourced from the Data-Smart Life & Annuities: Embracing Chat GPT & AI for Insurance Transformation Panel at Equisoft’s ITC Life Insurance Leaders’ Lab.

Author

AiUTOMATING PEOPLE, ABN ASIA was founded by people with deep roots in academia, with work experience in the US, Holland, Hungary, Japan, South Korea, Singapore, and Vietnam. ABN Asia is where academia and technology meet opportunity. With our cutting-edge solutions and competent software development services, we're helping businesses level up and take on the global scene. Our commitment: Faster. Better. More reliable. In most cases: Cheaper as well.

Feel free to reach out to us whenever you require IT services, digital consulting, off-the-shelf software solutions, or if you'd like to send us requests for proposals (RFPs). You can contact us at [email protected]. We're ready to assist you with all your technology needs.

© ABN ASIA