Davide Burlon, head of Actuarial Analytics at Generali, talks to Andrew Hobbs about how machine learning, the Internet of Things (IoT) and other emerging technologies are shaking-up the data-centric insurance industry.
As a data scientist who enjoys testing his analytical skills against rivals on data modelling competition platform Kaggle, Davide Burlon’s insights on machine learning and the IoT were always likely to be challenging. Internet of Business was keen to find out from him how innovations in the field will shape the insurance industry.
Burlon started out working in Generali’s retail pricing team, but soon recognised the need to explore the many ways in which machine learning could transform the industry, and meet the future needs of insurance providers.
This drive to make insurance into a model-driven business, open to the advances in machine learning, gave rise to the Actuarial Analytics team at Generali. It pools resources from the pricing teams, but maintains a policy of independently inspecting anything which is not a standard generalised linear model (GLM).
While GLMs are reasonably accurate in insurance terms, they are based on pre-assumed statistical distributions and other limiting factors. This makes other models, particularly those in the vanguard of neural networks, worth exploring.
Internet of Business: In what key ways is advanced data analytics disrupting insurance? What are the benefits for providers and, ultimately, for customers?
Davide Burlon: “I’m not convinced that analytics is disrupting anything in insurance – in the sense that disruption is a radical and fast change that reshapes a big chunk of the market within a short timeframe! At least not yet. It might be that actuarial and pricing teams have always been data-intensive teams, so new methods are strengthening or complementing capabilities which were already there.
“In my opinion the change which is taking place has more to do with the working mentality, the approach the insurance system has to distilling relevant information from data quickly, rather than discovering anything new.”
Can emerging technologies such as machine learning be easily applied to legacy systems, or are insurance providers having to rethink traditional models?
“There is huge work still to do in going from research to development, if you think of pricing and actuarial teams as a sort of R&D within insurance. This is also why I strongly believe there is a need for separate, dedicated teams that come from the actuarial world and are dedicated to finding modern statistical solutions.
I find the most pressing issue is the time to market, so it’s vital to allow for the continuous adaptation of the models, rather than finding the best-performing model statistically.
How is autonomous driving – and connected vehicle technology more generally – affecting insurance?
“The idea is that the more we remove human fallacies from the equation, the fewer the accidents. This is obviously everyone’s hope.
“From the insurance perspective, there is the additional cost component which will affect customers. Advanced driving-assistance systems (ADAS) typically cost much more than traditional equipment, so it’s still unclear what the impact will be on insurance premiums.
“Will the reduction in claims outweigh the increase in vehicle production costs? I believe the reduction in bodily injury claims will, thankfully, be more relevant than the severity of material damage costs, but that is just a theory.”
Do our increasingly connected lives, buildings and cities have implications for the sector too?
“I see two trends here. The market penetration of wearables might have a big impact on health claims by fostering healthier lifestyles, as well as prevention by detecting health issues and risks, before they become serious problems.
“The implication of the IoT in the home is still a big question mark, at least for me, while I find smart cities extremely interesting.
“A sneak peek of what the world biggest and most modern cities might look like one day is offered by crowd and services management at international airport hubs, like the next Daxing International or Dubai World Central, where data analytics plays a pivotal role in moving people vertically and horizontally in the most efficient way.
“Airports are somewhat easier versions of cities, because the who, where, and when of each passenger is determined a one point in time: the booking.”
- Read more: Amazon booms as high street withers. Next up: robots and insurance
- Read more: AtYourGate: Mobile app shopping takes off at airports
Are insurance providers that are exploring AI risk analysis doing enough to mitigate against the potential prejudices of so called ‘black box’ systems? Can this lack of transparency be overcome?
“Yes and no. I do understand the need of everyone in the value chain to comprehend the creation of a premium, but I think the prejudice speaks more to the mistrust of the customers regarding data protection.
For some reason, the population at large is mostly okay with giving away huge amounts of data to tech giants overseas, while at the same time we are asked to explain in laymen terms how a tariff is calculated.
“The two situations do not even compare in terms of magnitude, and yet insurance companies are held more accountable. Regulation is key in gaining public trust, and GDPR is a great opportunity to certify a levelled playing field.
“Insurers also need to step up their game and rethink how we sell our products. As customers, I think we are not be very interested in understanding how a neural network works if we know premiums are fair, discrimination-free, as low as the market allows, and I believe in the product I bought.”
Can the security, transparency, and decentralisation fundamentals of blockchain be successfully applied to insurance? Are there obstacles to overcome first?
“I see certainly a huge potential in claims handling and fraud prevention, but I’ve yet to understand the technology enough to determine whether a distributed insurance industry could survive big shocks and be profitable in the long run.
But this is an exciting time if you’re data passionate, because insurance companies are embracing new methods and technologies which have already proven successful in other sectors.
“From the business perspective digital transformation also forces us to re-think how we engage we customers, and the opportunities are massive for those who will be smart enough to ride this wave.”
- Read more: IBM, Marsh launch blockchain-based insurance network
- Read more: Banks’ use of AI, blockchain, cloud is systemic risk, warns industry
Internet of Business says
While Burlon makes a fair point about transparency in insurance premiums, fears about the risk of biases being automated by AI systems in the sector have been raised in Parliament.
In 2017, Lord Clement-Jones, who chairs the UK’s Parliamentary Select Committee on the economic, ethical, and social implications of AI, said: “How do we know in the future, when a mortgage, or a grant of an insurance policy, is refused, that there is no bias in the system?
“There must be adequate assurance, not only about the collection and use of big data, but in particular about the use of AI and algorithms. It must be transparent and explainable, precisely because of the likelihood of autonomous behaviour. There must be standards of accountability, which are readily understood.”