AI is not new, and most agree it will shape the insurance industry well into the future – but how can you maximise the value from it today? In the latest InsurTech Rising Bytes podcast series this expert panel explore where AI will have the biggest impact in the value chain immediately and what the near-term barriers are.
Moderated by Steve Pretre, Venture Partner at World Innovation Lab, with Jeremy Smith, COO at Risk Genius, Adrian Rands, Founder and Chairman at Quantemplate, Jerry Gupta, SVP at Swiss RE and Alex Baldenko, Lead Data Scientist at Mass Mutual.
The panel got started with the contentious question of “Is AI going to cause a mass extinction of jobs in the insurance industry by automating the insurance value chain?”
Jerry Gupta
Jerry kicked off,
“In the long term yes, but in the short term – probably not, and the panel agreed, adding that not only will some jobs disappear, but many will change and improve as well. This in turn is likely to lead to an increase in job satisfaction as AI takes on the majority of lower-mid level tasks, allowing people to add more value where they’re great.”
“It is not a case of demolishing the industry as we know it now, it’s making individuals more focused on what matters: understanding risk, customer service and delivering a better product.”
One case study to back this up was mentioned by Adrian and played out in much the same way as is now predicted, but almost 40 years ago. In the 1980s, he explained, the introduction of the fax-machine turned the insurance industry on its head. Roughly 10% of the headcount in the broking firm were working on the telex machines, and the fax machine allowed everyone to communicate with their clients, so there was some expectation that there would be mass redundancies.
“What actually happened was most of those telex staff who were familiar with what the industry was, moved into client facing roles, account executive roles and providing better service to clients and the companies themselves all grew substantially… It is not a case of demolishing the industry as we know it now, it’s making individuals more focused on what matters: understanding risk, customer service and delivering a better product.”
Thinking about AI in the insurance value chain
So while there will be a significant upheaval, the insurance industry shouldn’t be concerned about being replaced in the near term.
The panel then moved on to explore where AI will have the greatest impact – and comparing the differences between the “future-looking” scenarios to what can be achieved in the next six months or year.
Steve put it to the panelists to think across the spectrum of insurance; customer acquisition, underwriting and pricing, claims processing, risk mitigation and loss control.
Alex Baldenko
Alex gave his perspective from the life insurance side, and rather than choose one part of the value chain he emphasised that,
“The key to this is to put Machine Learning (ML) at the core of a business problem, as opposed to considering a separate research initiative.”
“You don’t need AI down at the bottom, you just need software. And humans still need to be at the top with the most complex tasks.”
However, where Mass Mutual are putting their focus in the near term is in customer acquisition – routing leads to the appropriate channel and agent to create a mutually beneficial experience for potential customers and their own agents.
Alex continued,
“Further down the chain in underwriting, our mortality risk model was built using our large data asset and in the near term is reducing time to approval to such a degree that the “policy not taken” rate has reduced by 30% since we started. Long term we hope to see a significantly reduced claims experience.”
Jerry agreed and highlighted that using AI is not just for cost-reducing measures- even in the short term- it is to create a way of making better and faster decisions.
Jeremy Smith
Adding to this, Jeremy reiterated that,
“AI will have the biggest impact across the board, wherever some degree of decision making is required.”
He went on to elaborate that,
“You don’t need AI down at the bottom, you just need software. And humans still need to be at the top with the most complex tasks. Where AI comes in is in the middle part, where some degree of judgement is needed.”
While where AI will have the biggest impact remains subjective then, Adrian believes the most immediate lies with data. Indeed, 2018 seems to be the year of data; some estimates put the amount data created each day at 2.5 quintillion bytes and since the new GDPR laws came into effect, companies have a serious task to protect, utilise and make the most of the data they collect. But it can seem an insurmountable task.
What are the near-term barriers to delivering on the promises of AI?
The final question the panel turn to is what might be standing in the way of insurance carriers when they try to kick-start their AI strategy. It’s an interesting question, which raises debate wherever it is asked, but Jerry opens with, “ignoring the conventional wisdom (though it is right) of available data and talent, two of the biggest challenges are integration and scaling internally, and explainability.”
Adrian Rands
Adrian added that usability and user functionality are absolutely paramount for successful adoption of the technology,
“A big barrier traditionally is the inability of the industry experts to be able to command and communicate with the technology itself. “
The last barrier, which may also be the most frustrating of all, was simply “time”. While there may be capability, interest and funding, many AI solutions will take time to bed in, be effectively tested and tweaked and for staff and managers to learn to use the outputs it can provide. The panel was concluded on the thought, “it’s clear that now is AI’s time – it just needs that incubation period, in order for the eureka moments to happen.”
Leading pioneers in the InsurTech space and ground-breaking new start-ups will join together at InsurTech Rising Asia in HongKong this October 22 – 23 2018. Check out the full agenda here.
Featured Image via Freepik
