Mankind’s mistrust of robots has been well documented through many years of dystopian science fiction. Futures filled with cold, objective automatons have been a mainstay of our screens for almost a hundred years. Even now, with robots and artificial intelligence (AI) beginning to form an ever-increasing part of our lives, suspicion remains. The British Science Association released the results of a survey in 2016 that found 60% of Britons thought the rise of AI would lead to fewer jobs within ten years and 36% of respondents believed it poses a threat to the long-term survival of humanity. However, the superiority of machines over humans is becoming more and more pronounced; chess grandmasters have long since fallen behind their artificial counterparts and recently the ancient Chinese game of Go, believed to be beyond the realms of artificial computation, fell to the mastery of machines.
(Re)insurance – an industry that revolves around data collection and analysis – will reap significant rewards from putting aside these fears and embracing AI. Incumbents, and start-ups, have already begun to explore such technologies and those companies that look to retain the status quo are likely to be left behind. Claims settlement is an area that has seen the greatest advancement in AI use and capabilities. Lemonade, the US-based start-up, was able to settle a claim in just three seconds in January 2017, using its bot “AI Jim”. Its AI reviews claim details, checks policy documentation and runs anti-fraud algorithms.
Established insurer Ageas is also leading the way with this technology and has begun using AI in its motor claims management. The automotive AI software the company uses is able to analyse images of damaged vehicles, assess what repairs are required and arrive at a claim estimate – all in a matter of seconds.
The uses of teh tech are not just limited to AI claims management. With the rise of telematics and connected self driving cars, more and more driving data is being collected. It is estimated that modern connected cars produce, on average, more than 20 gigabytes of data per hour. The amount of data being collected is quickly surpassing the ability of humans to analyse it. Octo Telematics, a leading provider of telematics products for insurance companies, has recently partnered with Cloudera, a machine learning and analytics platform. Together they are aiming to forecast trends and driving habits to better understand motor insurance risks. Their results may also be used to help identify and prevent types of insurance fraud, such as staged crashes.
Natural Language Understanding
Further uses for AI are being developed in the field of natural language understanding (NLU). Insurers are now able to use machines to search through online reviews, social media posts, blogs and online forums in order to extract information about policyholders that may be used to assist in risk assessments. Searching for this sort of data manually would be too labour intensive for an ordinary underwriter, making it effectively inaccessible without the use of NLU.
Concerns do exist around the ethics and morality of AI as an underwriting tool. It is possible that a machine-learning algorithm, when analysing the ever-increasing policyholder data that is being collected by insurance companies, could issue two, seemingly identical policyholders with drastically different motor insurance quotes. Underwriters will need to understand the computer’s ‘rationale’ for its rating and transparency around rating factors will be vital. For example, if Underwriting AI software deems university-educated policyholders less likely to have a crash than those who did not go to university, will insurers be able to charge the latter group a higher motor insurance premium?
The (re)insurance industry is on the brink of a seismic change in the way business is obtained, risks are underwritten and claims are settled. Machine learning and artificial intelligence are already being used by industry forerunners to speed up data-processing and analysis, allowing skilled staff to spend more time analysing the machines’ results, and less time doing the leg-work to reach the results themselves. It is unlikely that mistrust is going to hinder the march of robots towards ubiquity throughout the insurance value chain.