Pioneers from the 50s dreamed of fabulous machines that could imitate, replace and complement human intelligence—technology that bears human senses, reasoning and even empathy. Today, artificial intelligence (AI) bridges some of that gap and was adapted and built on a set of sciences and technologies that have made very interesting advances in order to enhance business across many industries.
Insurers, for example, began designing their marketing and pricing strategies, as well as streamlining the underwriting and claims processes, with the help of AI. When we talk about insurance AI, we are referring to a set of technologies and features that include: advanced analytics, machine learning, neural networks, intelligent automation, robotics, extended reality, text analysis and natural language processing, image processing, emotion recognition, IOT and big data (among others).
The value of AI lies in the ability to learn patterns, predict and classify, thereby projecting scenarios that can be analyzed to prevent and measure latent risks. If we complement this with technologies to automate decision making, we enter a context limited only by our own creativity to develop use cases.
Today we’re going to highlight some of the scenarios in which different leading companies in the global market are working today.
Pricing: Through predictive models (with algorithms such as random forest, linear regression, xgboost, etc.), we can provide insurance premiums in a more dynamic and precise way. More specifically, they can be personalized according to driving habits, geographic area and commute distance. To the traditional price-setting variables, a new set of variables are added to improve the profitability of the portfolio. These variables depend on the company’s own needs/capacities and can range from competitors’ prices to the policyholder’s traffic record, driver’s license age, credit score, as well as external data systems and sources. The interesting thing here is the dynamism in determining the price; the models change based on data inputted over time, then recognize patterns and adjust the rate autonomously.
Customer Service: Using real-time emotion and voice recognition algorithms, we can redirect calls from customers with serious problems to more experienced employees, directly impacting customer retention. In addition, the use of bots—real cognitive agents—has already demonstrated positive returns in productivity and efficiency when managing large volumes of calls, chats or emails, and their ability to operate 24/7 solves a large portion of customer queries and complaints.
Subscription: Predictive analytics helps us assess risk for each potential customer or insurable object, allowing low-risk policyholders the possibility of obtaining better rates. This can be done through grouping techniques, as well as considering individual and combined scores for different insurance risk classifications. Analyzing and processing images can further speed up the inspection process. And by using the person’s information, both recent and historical, private (such as smoking or having a heart condition) and financial, we can complete the person’s and policy’s onboarding in a matter of minutes, and in a completely digital way.
Segmentation and Sales: With the use of neural networks—learning based on decision trees, classifiers and big data technology—we can improve the segmentation and profiling of our clients and leads. We can understand them better through their buying preferences, interactions with the company’s contact points, and the use of data from external sources (eg: social networks). This can provide us with vital information for cross-selling/up-selling, providing better purchase recommendations, and in general, improving the customer experience. Another feasible example involves detecting significant internal relationships between clients and leads (father-son, spouses, partners, etc.), so that the treatment of claims and quotes can be better positioned for them.
Acquisition and Channels: We can use regression analysis, decision trees, classifiers and clustering to identify prospects and prioritize leads, improve marketing campaigns, and in turn, assign the best agents based on customer preferences. We can also predict agent’s probability and propensity to sell and identify more interesting target areas and markets to develop.
Retention: As we know, client retention is always cheaper than acquisition, which is why many companies are using structured/unstructured data, recurring neural networks and the combination of predictive models to identify at-risk policies and the likelihood of losing a customer. This enables the execution of early retention strategies.
Claims: Through digital expertise (photo and video, image recognition and fraud prediction models), we can reduce average claim resolution time to mere days, and reduce the cost of damage verification and claim assessment by up to 60% in insurers, both in the auto segment and in other risk areas. Then, artificial intelligence can identify damaged car or home parts, search for them in a cloud catalog and calculate the replacement cost. In addition to reducing administrative costs, the impact on customer experience, again, is invaluable.
Fraud: Applying predictive models in this process would not only allow us to focus human effort on most probable fraud cases (reducing false positive cases and increasing management success), but also make it possible to change the compensation approach. Combining an experienced professional with a good fraud treatment model, claim resolution time can be reduced to seconds or minutes, using automation to deposit the compensation or complete the repair order at the nearest workshop, just moments after the complaint.
ART: Many occupational risk companies are moving their prevention model to one of prediction, believing that “if we are able to accurately predict an accident, we’ll be able to prevent it.” There are studies where the analysis of a dataset (more than 112 million safety observations and the 15,000 incidents/accidents associated with them) has made it possible to predict incidents before they occur, with high levels of precision.
As we can see, the applications are endless, and our list only scratches the surface, since they depend on an individual company’s strategic focus and the business cases upon which they are built.
Finally, as with any project, it’s important to remember that success factors depend on the internal sponsor, long-term vision, strategic alignment, capacities to build and incorporate these types of solutions, alliances with experienced partners, commitment and constant monitoring.
