Artificial intelligence (AI) helps insurers to improve operations by breaking down traditional barriers to analytics. Modern AI can process and analyze data in various formats and structures, providing insights previously thought unimaginable. Despite concerns about data quality, claim departments hold a wealth of data, and AI is the key to unlocking its potential.
In an ideal world, claim data from first notice of loss to closure would be accurate, complete, and reported promptly. However, in practice, claim adjusters often encounter delays, incompleteness, and inaccuracies in data collection. Even when the data is in the system, it can be challenging to access and analyze it to identify important patterns. The ability to analyze data is further hindered by concerns about data quality, such as inconsistent coding, missing data, or inaccuracies. Many insurers mistakenly believe that their data lacks the integrity required for AI.
So why invest in predictive analytics if you have questions about your data?
The development of AI has made predictive analytics for insurance claims more accessible and feasible. The shortcomings in claim data can now be overcome with the flexibility and agility of current AI technologies.
The adaptability of AI technologies allows for the use of claim data in various formats, even if it includes missing fields or inaccuracies. Insurance analytics and AI are now able to handle most, if not all, shortcomings in claim data and still provide valuable insights on claim performance and trends that inform claim decisions.
Innovating with data in claims
Legacy claim analytics rely solely on structured data, which is stored in spreadsheets, databases, and tables. This data is easily accessible, sortable, and analyzable. It typically includes data such as claimant age, body part, nature of injury, location, and dates. This data is easily understood by both humans and computers. In contrast, unstructured data is not organized in a specific format and historically has been much harder to analyze and extract, as it includes images, text, video, and audio.
The most advanced systems today heavily rely on unstructured data to supplement structured data. One example is natural process language (NLP), a branch of AI that uses algorithms to interpret, understand, and generate language. In claims, NLP is applied to adjusters’ notes and other text data to quickly process and analyze all claims, enabling us to identify patterns in individual claims, as well as broader patterns in all claims.
The ability to analyze unstructured data, such as claim descriptions, adjuster notes, and other free-form text, provides a comprehensive view of the claim information. By utilizing NLP in conjunction with predictive models, claim professionals can be alerted to high-risk claims early in the process, before costs become problematic. When appropriate, these high-risk claims are triaged to experienced adjusters to apply cost-containment measures.
Additionally, the least costly claims can be identified, fast-tracked, and closed, freeing up resources for more complex cases. This claim triage process allows claims to be efficiently assigned to the most appropriate resources. By automating the claim process and relying on quantitative data, the time-consuming and error-prone manual review that often goes into claim assignments is minimized, resulting in increased efficiency and fewer errors.
In claims, unstructured data can be more informative than structured data. Structured data provides a partial snapshot of information that is easy to measure, but it fails to capture the evolving nature of the claim life cycle. Factors such as deteriorating injuries, disputes, attorney involvement, and discussions about treatment are not reflected in structured data but can be found in unstructured text data. AI now enables access to this data and offers insurers a path to gain insights from claim analytics and better understand the costs related to potentially expensive claims.
When AI is combined with domain knowledge, it brings a new dimension of value to the data that can reveal subtle yet meaningful insight into insurance claim trends as they begin to emerge. This allows claim departments to be more nimble and adapt to changes in their claim inventory. Improved data leads to more accurate and timely analysis, which leads to better, more efficient outcomes.
The increased specificity provided by AI allows claim departments to gain a more detailed understanding of the specific factors that impact their claim performance, rather than relying on general industry trends that may not reflect their experience. Instead of reacting to changes, claim departments can proactively adapt to shifts in trends and thoughtfully intervene as changes begin to emerge in the data, both structured and unstructured.
One of the added benefits of implementing AI in claims is the ability to improve overall data quality over time. By comparing unstructured data to structured data, AI can identify data that appears to be inconsistent with known trends. For example, obesity rates in claim notes are often underreported compared to the general population. Through implementing AI, claim departments often uncover gaps and inconsistencies in their data collection processes. Through continual development, data quality can be improved over time, resulting in more accurate insights and more precise predictions through predictive modeling.
By utilizing AI in claims, insurers can access data that was previously unknown to them, providing new insights into claim performance and giving them an edge in a data-driven market. To fully leverage the potential of claim data, insurers should seek out a partner with expertise in data science, a deep understanding of the insurance industry, and the ability to stay up-to-date with rapidly changing technologies.
This expertise has been brought together in Milliman Nodal, the analytics platform specifically built for the insurance industry and designed with actuarial science and claim operations expertise as an end-to-end solution for insurers. Companies that have deployed Nodal reported an average cost savings between 5% and 15%.
This article was originally published on June 18, 2021, and has been updated.