A big impact for simplified processes
The financial services industry has been inundated with reports about how technology is going to influence the insurance industry in the future. However, while a lot has been said about this growing influence, very little has been said about the actual impact that this influence will have on the insurance industry.
This topic was discussed in detail at the 5th Annual Spectacular Insurance Claims Conference where Tony Tsuen, Head of Technical Underwriting at Discovery Insure, said that the impact on the industry will be bigger than many expect.
Get the right skills
Machine learning will have the biggest impact on the industry as it will simplify processes and eliminate human error.
“For companies to get the best output from any machine learning methods used, they need to employ people with the right skill set, such as data scientists, statisticians and programmers. Employees with these skill set will be able to analyse data, come up with algorithms and successfully apply machine learning techniques according to the company’s evolving business needs,” said Tsuen.
He added that companies also need to invest in the right infrastructure, technology and software. Open source languages such as Python and R include packages that are often used to complete the machine learning steps outlined in the next section.
The data collected by companies is then processed in order to remove any duplicates, inconsistent or incorrect data and outliers. The data cleaning process will be applied repeatedly until the data scientist is satisfied that the data is ready to be used.
Data exploration and visualisation will enable the data scientist to represent the data visually in different ways. Based on this, they will be able to determine the limitations of the data and what they can use the data for.
Important algorithms
Tsuen pointed out that machine learning algorithms are applied to the cleaned data in order to learn from it. Generally, machine learning algorithms start as simple models and through hundreds or thousands of small iterations and layering become more powerful. Any new data that comes into the system is ingested by the model and improves model performance.
“The resulting model is applied to new data sets in order to predict the outcome or to complete a given task. This model can then be used in every day work by companies to predict outcomes with greater certainty. We have found that with more than 75% of companies investing in big data, machine learning is expected to increase in popularity in the next five years. In addition, it is predicted that 85% of customer interactions will be managed without any human intervention by 2020,” said Tsuen.
Lagging behind
While this is ground-breaking news (which may be scary to some), Tsuen pointed out that the insurance industry is still lagging behind, with only 1.33% of insurers investing in machine learning compared to 32% in software and internet technologies.
“There is massive potential for machine learning to be applied to different areas of the insurance business. A 2016 study done by SMA Research looked at the areas that had potential and this study indicated that 56% of new business underwriting in short term insurance could be achieved by incorporating some machine learning into the process. In addition, research has found that 74% of consumers would be happy to get computer-generated insurance advice enabled by AI and in addition, millennials are twice as likely to buy insurance online instead of dealing with a local agent,” said Tsuen.
Industry application
Looking at life insurance first, many companies are automating their underwriting processes.
US start-up Lapetus Solutions Inc (LSI) has designed a machine learning programme that uses a single photograph to underwrite life insurance policies. Using only a selfie, the algorithm can determine gender, age and body mass index to predict life expectancy.
Machine learning will have an impact on the short term insurance industry in the future. This may come in the form of improved modelling.
“Machine learning is currently being used to enhance traditional modelling methods. For example, machine learning algorithms are able to capture trends within the errors of traditional generalized linear models (GLMs). This is possible as machine learning algorithms search a far wider and deeper space than traditional models. As a result, this reduces human bias, increases accuracy and extends the possible tasks that companies can complete as they can build more complex models. In addition, ML tools enable companies to find better correlation between factors affecting risks at a deeper level,” said Tsuen.
Enhanced results
With all the extra information gathered, insurers have a far more holistic view of the client and can charge the client accurate premiums commensurate of their risk.
“An example of machine learning is natural language processing (NLP). This is a system that can be designed to immediately assess risks, calculate pay outs, automatically complete claims audits, perform administrative tasks, and more. In addition, some companies use machine learning as it can give prospective clients a score to determine their risk profile based on the company’s risk appetite,” said Tsuen.
While technology can exponentially benefit a company, it comes at a price. In future, insurers will have to sit and calculate what their return on investment would be when it comes to technology.
Editor’s Thoughts:
The tricky thing when it comes to this is that insurers cannot ignore technology adoption. Clients are looking for innovation when it comes to products, systems and processes. If an insurer is not offering it, their competitor probably will. Please comment below, interact with us on Twitter at @fanews_online or email me your thoughts jonathan@fanews.co.za.
