Every New Year brings with it a host of predictions of how our world will change in the next 12 months. For 34 staff at Japanese life insurer Fukoki Mutual Life, recent forecasts that 2017 would be a big year for artificial intelligence (AI) doubtless now seem particularly prescient. Replaced by a Watson-based AI solution for the payment of claims, its hard not to think these unfortunate employees will prove to be the canaries in the mine for many more financial services workers with replicable skills.

Whether hysteria or truth, there is a palpable sense that AI is gaining momentum right now. So too is the conception that technology is moving from enabler to destroyer of jobs, livelihoods, even society itself. However under- or over-cooked such talk is, it is impossible for senior insurance executives to ignore the potential savings AI can deliver to shareholders.

AI’s advance is not showing signs of running out of steam any time soon. IBM’s Deep Blue beating Gary Kasparov at chess in 1996 was a landmark moment, showed AI’s power in a single situation. The success of the same company’s Watson in beating humans in US TV quiz show Jeopardy in 2011 showed its ability to outperform its flesh and blood masters in multiple different comprehension situations.

The ability to interpret multiple problems and have the processing power to apply the process to the colossal amount of data now available means AI can engage with an increasing number of tasks formerly considered thought-based, delivering cost savings and efficiencies that go beyond the operational and into the realms of both underwriting and deeper understanding of customers.

It is well understood that businesses of all sorts can benefit from improved customer experience through speech analytics, decision trees and personalised interactions based on that individual’s data profile.

In the realm of insurance underwriting an AI system like Watson can take further intelligent steps. By consuming considerably more data about an individual and their circumstances, it is possible to get a more personalised risk for the individual, business or scenario being insured.

We are not just talking about all the data on individuals that’s floating around the web. Nor the internet of things inputs that will increasingly come into play, such as in domestic flooding situations. For example, smart white goods will detect deteriorating performance, speed damage detection, help identify what needs repairing and cut the cost of determining liability. What cognitive computing does is enable pretty much any data set or approach to be introduced into the mix to increase the sum total of the knowledge of the AI brain.

AI adoption is more advanced in the field of medicine than insurance. Just this month a team at Stanford University unveiled research showing AI performing as well as trained doctors in identifying skin cancers. Until now, variability of smartphone images caused by zoom, angle and lighting have rendered them unreliable for identification or classification purposes. But an AI data approach that uses 1.4 million images has achieved 72.1 per cent accuracy, compared to scores of 65.6 and 66 per cent accuracy by two qualified dermatologists. This new step potentially paves the way for melanoma diagnosis by smartphone photograph.

By generating lots of candidate answers and looking for evidence that these are the right answers, Watson can, with increasingly strong probability, answer complex questions.

IBM’s Watson has already ‘read’ more than 23 million abstracts of medical research journals on database Medline. That is an enormous amount of information – more than any one individual could ever read. The actuarial profession has nowhere near that much of a body of knowledge influencing risk, but while no actuary can have read every published paper in his or her field, an AI system like Watson can.

Insurers can use AI to absorb more risk drivers and expand the range of data areas used in underwriting. IBM is already working with global insurer Swiss Re, introducing a cognitive computing solution for its life and health reinsurance business. This will give it greater understanding of the risks it is taking on from the insurers it underwrites, making pricing more accurate.

Consumer-facing insurers should also be able to gain more insight on their customers through AI, as well as using it in day-to-day processes such as claims adjusting, administration and call centre operations. With deeper, more focused question and answer relationships underwriters can get more relevant information out of customers, and process it intelligently.

Information can be absorbed in many, many ways – the language used on a Twitter feed, on the phone, through a chatbot or in publicly available posts on Facebook can all, says IBM, be used to assess an individual’s capacity for and attitude to risk. This data won’t give a comprehensive answer, so a questionnaire will still be required. But it will complement the assessment, thereby making it more accurate than it would otherwise have been.

The extent to which AI will devour mid-range insurance industry jobs remains to be seen – Parkinson’s Law determines that the time these new systems frees up will soon get occupied by new tasks. But we can expect technology to bring us more efficient pricing of risks – and that has to be a net positive.

John Greenwood
Follow me

John Greenwood

John Greenwood is a multi-award winning journalist and writer. He is editor of workplace financial services magazine Corporate Adviser magazine, and was formerly deputy personal finance editor of the Sunday Telegraph. He has written for WSJ, the Sunday Times, The Guardian and numerous other titles. He is also the author of the Financial Times Guide to Pensions and Wealth in Retirement.
John Greenwood
Follow me