The industry of the future may bear very little similarity to the world you work in today. Here's why.
In recent years insurance has lagged other sectors in terms of adoption of IT. This technological lagging means that insurance as an industry is currently ripe for technological disruption. This disruption is occurring in three main areas, the understanding and pricing of risk, internal operations, and customer relationships. These changes mean that the industry is currently on the crest of a combination of technological advances which will utterly disrupt the current industry - a perfect storm. The industry is well aware of this, though few understand the scale of the looming disruption.
Big Data and Telematics
Traditionally insurance actuaries have estimated risk ratings from a limited sample of data, with clients forced into risk pools. Actual information on any particular client has been shallow and restricted. The new sources of data derived from areas like telemetrics, social media, or loyalty schemes will give insurers substantially larger amounts of higher quality data about their clients. Combining this with automated administrative, customer contact, underwriting, and claims systems, will enable insurers to assess each client individually in real-time and thus offer customized and dynamic policies.
Insurers will be able to use algorithms to automatically analyze the database for trends which can then be used to predict adverse patterns. For example; patterns of data which predict house fires, or link types of food purchase to types of sickness, can be discovered. The size of these data sets will also provide training sets for software algorithms to learn to handle non-routine cognitive tasks, hence starting the process of creating the AI software required for complex tasks.
This change to big data analytics will be one of the most important sources of future insurer competitive advantage. It will change the way insurers inter-act with their clients, the way they underwrite, and the way they structure their administration.
Insurers will gain a greater depth of understanding of personality profiles, buying trends and behavior. This will allow insurers to move away from simply descriptive (what happened) analysis and diagnostic (why it happened) analysis towards predictive (what is likely to happen next) and prescriptive (determining how to ensure the right outcome) analysis. Multiple scenarios can be run and likely outcomes role-played. Insurers will gain an in-depth understanding of why customers insure or don’t insure, and be able to individualize customer approaches so as to maximize sales.
Dynamic Insurance
Once real-time data feedback from telematics is added, the consequences for rate-making are revolutionary. The feedback from embedded devices will enable premiums to be set on an individual basis by software, and adjusted in real-time – ‘dynamic insurance’. This is insurance underwritten individually, with premiums dynamically set based on feedback from the telematic devices, e.g., car insurance which only activates when your car is on the road and increases or decreases based on driving behavior. Data analytics will reveal that a particular activity is risky for that kind of customer, and rather than excluding it, the insurer could offer a higher rate during that time.
House telematics can be linked to insurance company computers to report intruders, or fires, etc to accurate track risks in real-time. Houses will have finger print or voice activated security linked to doors or windows. Sensors liked to electrical wiring can shut down electricity flow if overheating and a potential fire is detected. Chips embedded in all house contents will allow their location to be tracked, even if stolen. This will lead to less claims and thus lower premiums.
Life or health insurance can be sold which is linked to wearable bionics which tracks blood contents and fitness levels, scan for signs of sickness, and alert clients to potential problems. Discounts can be given for health-related activities, warnings given if clients engage in unhealthy activities. Health specialists can decide which metrics need collecting, and data analysts can then start to analyze the flow of data to both feed alerts to doctors and to ascertain if trends can be found which predict adverse changes before they occur. If a client falls sick the health bracelet will tell the insurer that a medical emergency has occurred, and the insurer’s computer can arrange emergency assistance, analyze scans, and make payment of health bills without any need for patient involvement.
It will also allow detailed data analysis so that deep understanding can be gained around the relationship between client activities and well-being and sickness. The analysis of data from patients should discover trends which predetermine sickness and enable doctors to call in and treat patients before they get sick.
For example, UK insurer, Aviva, has combined data from a range of non-health sources, like shopping or online behavior and has found that these can predict future health outcomes nearly as accurately as blood or urine tests.
Currently insurers use actuarial based statistical algorithms based on compiling past data and event occurrences to forecast annual event probabilities. The new approach will be based on the structural drivers behind events as well as any conceivably related data. By being able to examine real-time structural data on casual factors, insurers will be able to price risk very finely. They will be able to offer clients attractive premiums if they meet certain behavioral conditions.
Internal Administrative Systems
Increasingly all insurance customers will be underwritten, and dynamically, rather than just at policy inception, or renewal. This is only possible cost-wise if customer interaction and internal administration is computerized, so the marginal cost of reacting to data is near to zero.
The looming transformation is thus as much about internal software and management systems as data. Current versions of automated analytical underwriting and claims systems have been shown to have the ability to both increase operations speed, often reducing claims processing times from months to minutes, and to cut the cost of underwriting and claims processing by up to 100x.
Big data is also starting to revolutionize risk assessment, because its intensive nature means that patterns can be found which are not visible when assessing a sample data set. For example, in health care, the digitalization of millions of client medical records has allowed software to compare each client’s individual symptoms with aspects like their genetics, their family background, their gut bacteria, or environmental factors, to create optimal and personalized treatment plans, and individualized medicines, all with minimal human oversight.
Similar techniques can be applied to customer relations, or to sales data; pin-pointing which marketing styles are important for each segment of the market. How, for example, do higher risk members of the white, female, suburban, SUV-driving group, differ from higher risk members of the white, female, city-center, non-car-owning group?
Client groups will no longer have to be made based on gross characteristics like gender or age, but on actual causal factors. Individuals from groups who currently face underwriting issues, like young drivers, will be able to prove that they do not individually possess the behavior patterns which make that group high risk.
Analytics will also allow customized client contacts. For example, it may reveal if a customer belongs to a group likely to lapse and at what point in time this lapse is most likely to occur, and then allow the insurer to create a personalized insurance policy contact structured in a way most likely to retain that customer.
Data-focused
The current focus of insurance management is on strict cost control, so that premiums can be kept competitive. These costs, however, are difficult to control and estimate, as underwriting estimates are created prior to policy issue, and are only updated at renewal time. Even then, there can be legal issues around substantial premium increases or policy refusal.
The new era of real-time big data via telemetrics means that insurers will have a qualitatively different quality of data on clients. They will be able to offer basic initial premium and then dynamically as client behavior reveals itself. Insurance needs to move from a simple closed analytical underwriting model, which has a set of equations, towards a more forward-looking model, which uses scenarios and structured cause-effect chains to give a deeper understanding of loss possibilities.
The main aim of data analysis will be the ability to extract useful customer insights. There are two main reasons for negative outcomes, (i) unreliable or unsuitable data, often provided by external sources and not properly incorporated into the organization’s own environment, and (ii) the analytics team missing an important component of data because they didn’t fully understand the business situation. Managers are needed who both understand data analytical methods and who understand the business.
The availability of big data is exploding yet the possibilities of its use have been little explored in the insurance industry so far. While some insurers have invested in some areas none have a vision of an integrated system, based on near zero marginal cost admin systems.
Data Sources
Current insurer client data is historical, shallow, and based on very few interactions, whereas new data will be deep, rich and real-time. Complexity will arise because underwriting data will need to be combined from as wide a range of sources as possible, many of which can be quite different; for example, minute-by-minute purchase data, retail data from loyalty cards, location data, text, online comments, blogs, and call center communications. Combining these diverse sources can yield unsuspected insights. For example; Woolworths Australia discovered that customers who drank lots of milk and ate lots of red meat had a significantly lower auto-insurance risk than customers who drank spirits, ate lots of pasta, and filled their petrol tanks at night.
The size and complexity of this data mean that analysts cannot visually purview data but will build AI systems to mine it for useful insights. Key skills will be big data analytic skills and data presentation skills, as well as deep knowledge of customer behavior. Existing data software suppliers may not survive the transition if they do not have skills in AI mining of large and disparate data volumes. Instead social media firms or consumer internet players, who do have experience in AI mining, will probably take over supply of data software to insurers.
One issue is that in large insurers, data is normally trapped into silos and locked behind access controls so it is virtually impossible to gain an integrated overview of reality. Data governance and ownership is also normally spread across an organization so that there is no centralized view of what data they have and how it is used, so that it is impossible for any creative new uses of the data to be envisioned. All these challenges mean that existing insurers will have to create entirely new IT, data flow, data visualization, and decision systems based on new management processes.
Big Data and Artificial Intelligence
A vital part of the response by insurance survivors has to be the creation of effective AI systems using adaptable learning algorithms. The use of superior AI systems will allow insurers to mine big data derived from telemetrics and business-ecosystems to discover activity-risk correlations which are not obvious to actuaries, and to cut marginal cost substantially. AI will thus be a key component of internal administrative systems, including their ability to respond to customers via auto-generation of emails or phone calls or social media posts, (using text and vocal recognition skills) as well as responding actively to telemetrics feedback.
A key aspect is that insurers will work as part of a wider ‘business ecosystem’ of related firms. Expertise in AI will allow insurers to integrate their products into the telemetric-based products used by other firms in their eco-system, enhancing these products and extracting vital data. One of the keys for insurers to become a customer-centric company is a focus on the integration of big data into client-facing activities, delivered via a radical reduction in per-service administrative costs.
Insurance as a Service
Linking insurer systems to telemetrics offers a range of exciting revenue sources for insurers. Feedback from telemetrics and accurate predictive indicators from big data will allow insurers to move from compensation after an event to advice and warnings on how to prevent an event. The aim of these can be marketed initially as being focused on reducing client risk, but the real aim would be to integrate the company into customer lives. An insurer could then brand itself as handling a customer’s lifecycle management. Insurers will change their business model from being one of providing insurance products to being one where they are data companies with real-time links to customers, specializing in data-based personal risk products and services.
