Tuesday, 08 August, 2023
Accelerated claims processing, efficient use of data assets, and more satisfied customers in the end – the collaboration of humans and Artificial Intelligence (AI) in the claims process can create significant added value for insurers. In the process, AI does not operate in a “black box” but is organically embedded into everyday work, in the guise of augmented intelligence.
About the Author: Dr. Wolff Graulich is Managing Director at the digitalization expert Eucon. He holds a doctorate in molecular biology and is an esteemed expert in the insurance environment and in cross-industry digital transformation.
Insurers planning to use AI and data analytics are looking for more than just process automation. The goal is also to retain knowledge and expertise within the company. In addition, it is crucial to ensure that experts do not waste their capacities on routine tasks, but can devote their time to complex activities. AI gives them this freedom by automatically performing a growing number of time-consuming routine tasks on their behalf.
Why AI needs experts
That man and machine are not competitors but actually work more efficiently through collaboration can be seen in the model of augmented intelligence. Humans are not excluded from the decision-making process; rather, they make targeted decisions as soon as a claim requires more than simple routine processing. AI merely provides them with supporting recommendations for action based on analyses that have been carried out.
When introducing AI tools, insurance companies must already decide where the human mind offers the greatest added value and which tasks can be performed more efficiently by “colleague AI”. Employees are looking to add meaningful value through their work and seek to bring their experience and skills to the table. AI, in contrast, is more efficient when it comes to repetitive routine tasks and the preparation of sound decisions to be taken by humans.
Even though AI is expected to work more and more efficiently in the future, humans will remain crucial in the process, not just in the application. AI is like a machine that requires manual maintenance. It needs up-to-date data, feedback and training. By performing professional quality assurance before application, problems such as overfitting can be avoided. The discussion about generative AI has shown that quality monitoring is indispensable. This is already built into augmented intelligence.
Continuous feedback improves the technology
The best way to explain how interactive machine learning works is by looking at a real-life example. When checking a motor vehicle claim, the AI uses learned models to analyze data points that are more prevalent in fraudulent claims. Indicators of anomalies to be reviewed can include the vehicle age and mileage or the presence of specific types of damage.
Likewise, information as to how long the vehicle has been in the possession of the current owner helps with the classification. At the same time, AI identifies other aspects as unproblematic. The results are displayed in a transparent and clear manner that allows the claims managers to check them at their own discretion and based on their wealth of experience. They can either analyze only the conspicuous elements in more detail, or check everything in equal measure. With sound AI performance, claims managers can eventually be assured that the items found to be innocuous are indeed highly unlikely to require in-depth scrutiny. This gives them more time to focus on the points of concern and to track down potential fraudulent cases more efficiently.
In addition to the usual “re-training” of the AI, it is important for claims managers to provide feedback on the course of the process. Were there any faulty conclusions drawn by the AI? Should more data have been included? This helps improve the AI and further develop the data analysis process to produce more accurate results in the long run.
Transparent data processes – the basis for regulatory compliance
A transparent work process is indispensable where strict compliance regulations are in place. This is especially true for data analytics processes that support decision-making. In the insurance environment, AI is not permitted to work in a black box, where it autonomously makes decisions that are not comprehensible on the outside. It must be possible for human assessors to logically justify claims processing, if only in the event that injured parties dispute the decision. Errors in the process must be traceable in critical cases. Insurance companies should therefore pay close attention to the aforementioned criteria when selecting tools and service providers.
In addition, insurers that use data analytics must ensure that they retain control over their data. This applies to internal utilization as well as to collaboration with partners and service providers. For compliance requirements to be met, it must be comprehensible, both qualitatively and conceptually, how the data analytics process was carried out. In other words, insurers must be able to tell who worked with the data, when and how, and what measures were taken to prevent data breaches, for instance.
Outlook: What are the technological challenges facing insurers?
Today, insurance companies are already very well positioned in terms of technology. However, productivity gains from workflow support software are flattening out. The effective use of data offers them a competitive edge in this context, as it upgrades existing workflows and enables new process steps through early routing and process control.
The biggest challenge facing insurers is making their own and external data available in real time and moving and analyzing it within processes – that is, at the interfaces. Data must be available in usable form in the right places, both internally and also at partners and service providers. This is a complex task, both in legacy systems and in modern IT environments.
Insurance companies will have no choice but to support their business more with data analytics and AI in the future. Through the collaboration of man and machine, they can become more efficient, increase the satisfaction of employees and customers, and tap new value creation potential.
Written by Eucon Group