Optimising claims management with AI
September 2021 | TALKINGPOINT | SECTOR ANALYSIS
Financier Worldwide Magazine
September 2021 Issue
FW discusses AI claims optimisation management with Axel Linke and Sandra Krasny at INFORM.
FW: Could you provide an overview of how the claims management process has evolved over the last 10 years or so? What do you consider to be the key developments during this time?
Linke: For more than a decade, we have seen a massive transformation from paperwork to digitalisation when it comes to claims handling. Paper dossiers are increasingly being replaced by digital dossiers, and legacy systems are being replaced by more modern and convenient front ends. Also, many common business processes are now digital, leading to ever more automation, especially for standard claims. But all too often, these efforts are not embedded into a complete digital strategy for claims handling. This leads to inconsistent decisions and processes, like multiple service providers for bill reviews or prolonged settling times because of uncertainty on how to proceed. To embed all these isolated solutions into a streamlined digital experience for the customer and achieve significant cost savings along the way will be the challenge for the coming years.
Krasny: Advanced technology has been disrupting the entire insurance industry and we have also seen the first InsureTechs enter the market over the past several years. The potential of big data is tremendous, and insurers are now trying to make the best use of it. Those who manage to leverage their data, combined with technology as a strategic asset, can win new business, despite the extremely competitive and saturated market situation. Today we are seeing a strong focus on optimising the entire claims management process through automation. There is a long way to go, but the goal is clear: more efficiency, reduced costs and satisfied customers through increased service quality.
FW: To what extent have you seen an increased adoption of artificial intelligence (AI) among insurers? Is there a rising demand to explore AI technology and integrate it into the claims management process?
Krasny: The amount of available data is growing exponentially and efficiently managing this expanding pool is accompanied by great challenges. The industry is, at the same time, realising the increased value of their data in combination with AI and machine learning technologies. Those that can scale their IT infrastructure and take advantage of new data sources will move to an AI-driven future soon. There are numerous application areas along the entire claims management process and expectations are high. Nevertheless, AI is only as powerful as its underlying data.
Linke: There is a large demand for AI-driven solutions in the insurance industry. However, it is still somewhat unclear for exactly what purpose it will be used and how this is integrated into the whole claims-handling process. Currently, it seems many insurers are still in an experimentation phase rather than adopting and deeply integrating these technologies into their processes. AI technology is like a huge toolbox. To be successful, the right tool must be chosen for a specific task. Quite often, it is not the turbo stabilised pneumatic hammer drill that is required; rather, a simple screwdriver will suffice.
“Those that can scale their IT infrastructure and take advantage of new data sources will move to an AI-driven future soon.”
FW: In what specific ways can AI help insurers to optimise claims management? Where can it boost speed, efficiency and accuracy?
Linke: If we look at claims handling today, many small tasks have been automated or outsourced to more efficient service providers. However, one vital key to success is decision making in claims. Every claim requires a lot of decisions. Shall we hire an expert? Shall we send the bill to an external review? Do we need photos? Do we need a written statement? Do we trust the claimant? Can the claim be paid? All these decisions must be made by the claim handler, based on the facts at hand. But there is a lot of data around the claimant, the car, the policy and their history that a claim handler is not aware of. AI technologies can make use of this data within milliseconds, supporting the claim handler throughout the decision-making process with helpful prompts and tips. This, in turn, will result in quicker, cheaper and high-quality decisions. It is not only about complete automation, which is also very hard to achieve. A big advantage lies in teaming up AI with the human brain.
Krasny: The assessment of a claim can be a long and painful process for both the customer and the insurer. This is particularly true in the car insurance domain where a large amount of data is available. AI can improve the claim settlement part significantly. In recent years, insurance carriers have collected a huge amount of data and built up giant data lakes. This is fertile ground for implementing digital decision-making systems. Based on past and current data, the entire process from ‘first notice of loss’ (FNOL) to closure can be executed automatically. A substantial portion of tedious, manual work can be picked up through AI support, which, in turn, will lead to faster claim-handling times and reduced costs. As the insurance industry is still at an early stage of this digitalisation journey, there are a lot more application areas to tackle, such as pricing and underwriting, as well as internet of things and device monitoring.
FW: Is there a difference in how AI — and associated technologies such as machine learning (ML) — should be deployed to assess and process more complicated claims, compared to more straightforward, lower-value claims?
Krasny: What we currently see on the insurance market is an increasing amount of InsureTechs – primarily small, entrepreneurial companies – which focus on AI and mobile applications. Most of them cover property and casualty insurance business with standard and lower-value claims. This is a perfect playground for AI and its associated technologies, as you can fully automate the process from end-to-end. When you look in the areas of life, enterprise or commercial insurance, you will rarely find InsureTechs. The market is instead shared among the traditional international players. Here you typically have more complicated and expensive claims, where customers expect human interaction. Beyond that, a €15m fire claim with a burned building and injured people does not happen often enough that we would have enough data to even train AI. That does not mean that everything needs to be done manually, but deploying AI must be carefully considered and only executed in concrete use cases.
Linke: Segmenting claims into categories such as ‘easy’, ‘standard’ and ‘complex’ is one way to break down complexity. However, it is always possible that a claim evolves from one segment to the other because of new information. Also, highly automated claim settling requires a capable fraud detection system. Fraudsters quickly learn how to exploit the system in their favour. A good approach would be to define a long-term digital strategy, then identify the major modules of the claims-handling process and start integrating a system step-by-step. This is simply because the more complex the claim, the more complex the AI solution.
“AI is a powerful technology and we have been provided with only a glimpse of its full capability.”
FW: How would you weigh up the pros and cons of using AI in the claims sphere — particularly its impact on the human dynamic?
Linke: The potential of AI technology is huge. It can be seen as a threat, because it is now able to do things that were considered impossible for humans, such as aggregating large swaths of data and quickly drawing conclusions from it. There are certainly things to keep in mind when weighing the pros and cons of AI integration into claims handling. On the pro side, using AI can lower costs, increase quality and standardise business processes without oversimplifying. One topic that must be considered, however, is ‘explainability’. Insurers must be able to explain their decisions for regulatory reasons or in a lawsuit. Simply stating ‘the machine told me to do it’ will not suffice, which highlights the importance of taking a white-box approach to AI integration for decision making. Knowing all this, the best way forward for AI integration seems to be a team effort. Combining human intelligence, gut feeling, creativity, empathy and common sense with AI’s advantages, such as pattern matching and speed, is the perfect match.
Krasny: You can implement the best algorithm in the world, but it will not help if it is perceived as a black box and the addressed human does not accept or understand the outcome. The lack of transparency of AI, where we do not understand why or how a decision was taken by a predictive model, is probably one of its biggest weaknesses. Especially in the claim sphere, ‘explainability’ is crucial for situations where the insurer must be able to defend and document its decision. Nevertheless, the advantages outweigh the disadvantages. AI can predict what might happen in the future and make proactive decisions in the moment without human intervention. Whether AI is successful or not is a question of how it is used.
FW: Based on your experience, what best practices can insurers follow when implementing new AI technologies into existing systems and processes, to improve productivity and performance?
Krasny: Evaluate your current situation and define your target very precisely. The amount of data and technical possibilities will be overwhelming, and even if it is tempting to solve as many problems as possible within one project, we would recommend starting small and focused. Most companies have big expectations and underestimate that AI is only as good as its underlying data, since the output depends directly on data quality and availability. Best practice also shows that business and operations must work closely together. For example, when you ask IT to deliver information on a vehicle type and model year, they might tell you that they do not have this field. But the business knows that you can read this information from the unique vehicle identification number (VIN) using a decoder. Finally, for a sustainable and profitable deployment, the system must be constantly maintained and evolved, which requires dedicated resources.
Linke: The first thing to do is analyse the affected process and streamline it. Old business processes often have a lot of exceptions and special rules. An audit of such factors would be a good place to start. Be sure to involve claims handlers in the discussion from the beginning. There are always two realities – how things should be done and how things are actually done. The delta in many cases is surprising. Then, design the new process on a blank sheet of paper. Differentiate between the 80 percent standard, which have a high potential for machine-controlled automation, and the 20 percent exceptional cases. Solving the 20 percent efficiently will require the most amount of time and effort, but if this can be managed, a true digitalisation strategy can be achieved. This is also the part where humans and machines will have to work together to achieve cost savings.
FW: Looking ahead, what are your predictions for the application of AI in claims management over the coming years? In what ways do you expect the toolkit to evolve?
Linke: The adage ‘forecasts are difficult especially when they concern the future’ rings true. AI is a powerful technology and we have been provided with only a glimpse of its full capability. Therefore, we believe it will become increasingly valuable in the future, optimising an increased number of processes, and taking on a bigger role in vital everyday life decisions. The composition of staff and the required skillsets will also evolve with time. There will be a significant increase in IT and data analytics hires within insurance organisations. A good idea would be to start hiring talent directly from school or university and try to keep them on board so there is some consistency within the staff as this digitalisation journey unfolds.
Krasny: Things like smart home, wearable devices, telematics and mobile applications will lead to a continuous growth of data. The increased amount of information and improved analytical measurements will lead to new products and offers, such as usage-based insurance. Covering existing and new risks based on a unique personal situation and customer needs will no longer be just a niche product but penetrate the market. AI can assess in real time the associated risk, calculate a fair market price, and do an automated settlement in case the risk turns into a real claim.
Axel Linke has been a senior business consultant for the risk & fraud division at INFORM for six years. Focusing on the insurance industry, he has extensive experience in process optimisation and data analytics. Before joining INFORM, he worked for a large German insurance company for 13 years where he focused on workload distribution and workflow optimisation across all lines of business. He can be contacted on +49 (0) 2408 9456 5000 or by email: riskshield@inform-software.com.
Sandra Krasny works as a senior business consultant for the risk and fraud division at INFORM and is responsible for the Germany, Austria and Switzerland market. She has been involved in many RiskShield project implementations and has more than eight years of experience in fraud detection and process automation at many insurance companies. She can be contacted on +49 (0) 2408 9456 5000 or by email: riskshield@inform-software.com.
© Financier Worldwide