AI Gets Real

Not many years ago, artificial intelligence (AI) was the stuff of movies like I, Robot or 2001: A Space Odyssey, or of novels by Isaac Asimov or Philip K. Dick. It is now a reality: cognitive computing, machine learning, natural-language interfaces and related technologies allow computer systems to interact in a human fashion, make decisions and apply outcomes mechanically, to have the appearance of sentience.

The insurance sector is waking up to the disruptive nature of this technology.

Norman Black, Insurance Industry Principal Consultant, SAS

Accenture’s Technology Vision for Insurance 2017, findings of which are based on a survey of more than 550 insurance executives from 31 countries, clearly shows the insurance industry is well aware of AI’s potential. In all, 75% of respondents report they believe AI will either significantly alter or completely transform the overall insurance industry in the next three years.

Two-thirds of those polled say they expect AI to significantly change or completely transform their companies.

AI, BIG DATA AND IOT

Two related trends give AI meaning and context: big data and the Internet of Things (IoT).

Big data

Moore’s Law, which states that computer technology will double in power and halve in cost every 18 months, is not slowing down. Nowhere is this more obvious than in the field of memory.

This reduction in cost — along with new database technologies — enable the processing of vast amounts of data at terrifying speeds in memory and the processing of both unstructured data (think, among other things, text records, social
media postings and video) and structured data.

Since various sources estimate that anywhere from 70% to 90% of organizational data is unstructured, this is a significant development.

Internet of Things

There is plenty of data for those systems to work with, thanks to the increasingly connected nature of the world. Consider the average new vehicle, with its engine sensors, body and chassis sensors, telematics, global positioning system (GPS) functions, seatbelt sensors and tire pressure sensors, perhaps amounting to as many as 200 per vehicle by 2020. It is an object lesson in the potential that the confluence of big data, IoT and AI brings to the table.

BUSINESS PRESSURES
Against this background of rapidly advancing technology, insurers still face traditional business pressures, namely profitability and revenue. A KPMG survey released this year found the majority of polled industry chief executive officers predict top-line growth of less than 2%.

Traditional pressures are being compounded by these newer technology trends: 46% of the polled executives say they expect major disruption in the sector over the next three years because of technological innovation, and 36% name emerging technologies as their greatest risk.

It is not simply a matter of keeping up with the process and operational efficiencies that can squeeze more margin out of the business; it is about fundamentally changing the business from process-driven to customer-centric so that the standards are not being set by competitors in the industry. Sectors such as retail and travel are setting the customer experience bar ever higher.

Millennials — the first generation that was born into an always-on world — are now entering their 30s, and soon their 40s. As they become a prime target for the insurance market, they bring with them expectations of a frictionless customer experience. A 24-page insurance document surely tries the patience of someone who has grown up with oneclick shopping and recommendation engines — and the confidence that an alternative is a few clicks away.

Technology also brings the risk of disintermediation —the “Uberization” of the industry by business models that have been created without the shackles of traditional industry processes and norms. Direct and aggregation models are nimble enough to capture that Millennial audience, so much so that in many insurance markets, incumbents are buying out or investing in new “insurtech” start-ups, hoping to develop complementary services.

DATA IS CURRENCY

More frequent and positive engagements are the next step in the customer experience, and it is one that few insurance companies have taken. The contact between insurer and customer are few and usually negative; a renewal notice that
bills for premiums, or a loss claim.

But there are exceptions. Take, for example, a health company in South Africa built on the premise that it is more profitable, for the customer, the company and society as a whole, to reward healthy behaviour rather than punishing unhealthy behaviour with higher premiums. Through its business structure, the company logs 70,000 gym visits a day, has offered discounts on hundreds of millions of dollars of healthy food purchased at partner stores, and offered reward points for customers logging their workouts.

For the right incentive — and given the commensurate guarantee of its safety and privacy — customers are willing to exchange the personal data that is the foundation of an AI program. Yet 43% of the executives surveyed by KPMG say that they felt a lack of quality data was hindering their insights into customers.

That said, there are a growing number of channels to gather relevant customer information. Consider that clients of the noted health company carry the equivalent of a loyalty program card for use at partner organizations. Connected homes can also provide a wealth of information that has a bearing on risk and offers an opportunity to engage partner service companies. But perhaps the most compelling example of the use of data and AI in an insurance context is the automobile.

Insurance is one of the costliest ongoing expenses of automobile ownership, and one that could be strongly influenced by technological advancements.

The current regime — sex, age, experience, driving record, number of annual  miles and location — was once the best actuarial model available. Monitoring actual driving behaviour would be even more predictive of risk, and with the many vehicle sensors producing telematics that can be collected in real time, that model is now possible.

Still, it is also worlds away from the current customer renewal model. Integrating readily available public safety and traffic data into a system can guide the customer through faster or safer routes from points A to B. When telematics indicates a car component is
failing or in need of service, or if the vehicle is simply running out of gas, the system can guide the driver to an authorized partner service centre or gas station where he or she will receive a discount.

And in the event of a minor accident, the car can be self-reporting. Soon, customers are driving more safely because they know it will be reflected in their premiums — a win for the insurance provider, the driver, the health system and society in general.

READYING FOR THE JOURNEY

How does the insurance industry ready itself for the brave new world of AI? At the highest, most strategic level, companies must be prepared on three fronts.

Technology roadmap

It is not simply about buying new hardware and software. It is about having a technology platform that can nimbly accommodate changing business strategies and priorities, while ensuring control over data use, security, redundancy, replacement planning and appropriate provisioning, among other things.

Industry politics

Brokers tend to hold customer relationships in personal and commercial insurance; they are the ones who meet the clients and understand their business (both corporate and personal). If a company with bigger pockets can gather and analyze more information about the client and can know the client better, who should be the caretaker of the relationship? Should the insurer partner with the broker to help them gather the same intelligence?

Talent and culture

This should almost be considered an essential part of a company’s infrastructure. Some companies are setting aside resources for a “digital garage” — a sandbox for post-graduate data scientists to be given free rein to push the boundaries of how AI can be used.

A background in insurance is not necessary; in fact, it might even be a detriment to unbridled curiosity.

Data science is, however, a nascent field academically, producing fewer than a couple of hundred graduates a year in Canada. Using this very scarce talent efficiently and effectively is paramount to success.

Insurers need a platform that enables data scientists and analysts to collaborate and create, using the language and tools of their choice, but that also delivers the disciplines and structures needed to manage data consistently and deploy new AI solutions at scale across the organization

-Norman Black, Insurance Industry Principal Consultant, SAS