By virtue of the risk adverse nature of the insurance sector, the sector has not been considered to be at the vanguard in terms of the adoption of new technologies.
However, the recent announcement that Fukoku Mutual Life Insurance is making 34 employees redundant and replacing them with IBM's Watson Explorer artificial intelligence (AI) platform to calculate insurance pay-outs and, in so doing, increase productivity by 30% and save on salary costs in excess of £1 million per annum, is yet further evidence that the insurance sector has embarked on a journey of technological innovation from which there will be no turning back. Although it is not possible to predict where that journey will take it, one can state with certainty that over the next decade or so technology will transform the sector beyond recognition.
The nature of the insurance sector lends itself to automation. Insurers have always had to handle, analyse and process large volumes of data, which of course historically was effected in paper form. Initially, the sector’s automation activities have been focused on simple rules-based claims processing and automated underwriting, lead generation and advisory activities.
It is, however, the emergence of “big data” which, when combined with advances in machine learning, is accelerating the speed of transformation. Big data is the ability to analyse extremely large data sets to reveal patterns, trends, and associations, especially relating to human behaviour and interactions. Machine learning is a subset of AI and is the process of computerised detection of data patterns and analytic processing. It is through these advances, that insurers are better able to understand their customers, what is happening within the insurance markets and to develop systems that will be able to think.
The most obvious financial benefit to insurers of deploying machine learning will be in terms of cost savings, arising out of a reduction in the number of human resources required to be deployed.
The benefits are potentially much more far reaching: insurers will be able to use machine learning to do things smarter and faster. Claims handling will become increasingly more efficient, with claims processing times (for non-complex claims) being reduced from a number of months down to a number of minutes.
Analytics will also enable the swift identification and prioritisation of those more complex claims, which require more experienced adjustors to deal with them, thereby accelerating the speed with which they are dealt and minimising the risk of litigation.
Further, as machine learning will be more accurate and consistent than human resources, this will lead to a reduction in the number of claims being rejected, which rejection is overruled on an appeal. AI and machine learning will also be able to have a significant impact in terms of materially reducing fraudulent insurance claims, through a combination of modelling, rules, text mining and database searches. This should lead to lower costs and lower premiums.
Overall, AI and machine learning will enable insurers to have a much clearer view of risk – and consequently enable them to price risk with much greater accuracy.
As AI evolves, from “systems that think”, into “systems that learn”, the possibilities within the insurance sector are even more significant. One can anticipate that AI will be deployed to assist human resources in their activities. In addition, it is likely that AI ecosystems will be developed, which in turn will lead to an evolution from man-machine learning to dynamic underwriting, virtual assistants and robotic-advisory capabilities.
Although this “Brave New World” should deliver benefits to customers, in terms of lower premiums, greater efficiencies in terms of claims processing etc., there are a number of legal issues that need to be considered. By way of example, where AI is deployed to take decisions that affect individuals, the new EU General Data Protection Regulation will be potentially of relevance. The Regulation states that individuals:
“shall have the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her”.
There are exceptions to this rule: where an automated decision is provided for by the law (e.g. fraud prevention systems); or is necessary to enter into a contract; or is based on an individual’s prior consent. However, in the latter two cases, individuals will have the right to obtain human intervention to express their point of view, to contest the automated decision and receive justification of the automated decision.
With regard to the insurance sector and its use of AI and big data, decisions as to whether to insure an individual or approve or reject a claim, could be too complex and based on a too large number of data points, that it is not possible to give a justification of the automated decision.
To address this, in developing their AI capabilities, insurers will need to ensure that the decision making processes are structured in such a way so that it will be possible to track the reasoning of the decision. However, at this time, it is unclear as to what level of justification would be sufficient to meet the requirements of the Regulation. In conclusion, there is no doubt that AI will present the insurance sector with significant cost savings and efficiencies over the next decade. Big data and machine learning will revolutionise how risks are underwritten and claims handled. However, there are justifiable legal risks with the "artificial" element of AI, particularly when dealing with individual customers, and so human intervention cannot be made wholly redundant, yet.
About the Author Peter Dickinson is a Partner and Co-Head of the Business Technology Sourcing Practice at international law firm Mayer Brown. Peter specialises in advising on large-scale multi-jurisdictional outsourcing projects.