February 20, 2018

At LexisNexis Risk Solutions we have been using elements of machine learning and artificial intelligence (AI) in data processing and risk analytics for many years.

What we have learnt is that AI and machine learning require a strong data foundation to be effective. There are some important principles to follow, to truly reap the benefits of an AI-powered application. With this in mind, we wanted to comment on some of the key questions that come up with artificial intelligence: what exactly is driving the rise of AI, and why does society need it?

From a practical viewpoint, what is artificial intelligence? How is it different from machine learning? Where is the best place to get started?

In terms of where businesses have been deploying AI up to now in the enterprise, there’s been some interesting research recently from Forrester.

The analyst firm in its Global State of Artificial Intelligence Online Survey found that marketing and product management are the two leading areas where corporations have been investing in AI. When questioned about which areas of business are leading the adoption of AI, 46% of firms globally said marketing, followed by product management (40%), customer support (40%), the CEO or board (25%), risk and compliance (19%) and security (19%).

Forrester found that a rising number of data and analytics decision-makers claim to have unencumbered access to insights, effectively new insights being harvested from big data.

This trend towards higher and better business insight from AI is something we can expect to continue growing. The stimulus for this accelerated democratization of data will be embedding AI, big data, and IoT into analytics processes.

Product management and customer support driving the rise of AI for insurance

Empowered consumers for insurance are now on the move and they are ready to reward or punish companies based on a single experience. This behaviour, once thought of as being confined to the Millennial generation, is now widespread. It has been created by technology, the Internet, wireless and mobile devices, meaning that the consumer is now in control. No longer do they need access to a salesperson for information or pricing.

Just as we have Millennials looking for information in any context or platform they choose, older consumers have learned the same behaviour. They too are looking to customise everything they buy online, from an insurance policy to music streaming, to purchasing a car online.

Buying behaviour has now changed across the whole age spectrum and products need to be served up proactively. As a result, the challenge of managing today’s cross-channel personalised customer engagement is beginning to exceed human cognitive capacity.

There is now a widening gap between the traditional route for quoting and on-boarding customers, and the modern route. Companies that are not adapting will not be able to sustain the trajectory to stay competitive.

All of this is leading to AI-powered applications within the technology stack of insurance companies, creating efficiency, speed and smarter decisions related to pricing and personalisation.

What exactly is artificial intelligence and how does it differ from machine learning?

In a recent meeting sponsored by marketing service provider to the technology sector MRP, Steve Casey Principal Analyst at Forrester, commented that it is the rising volume and velocity of information that is really driving the demand for AI solutions.

Artificial intelligence can be described as a system with three steps: first comes the sensing (ingesting data into the system), then the thinking (the machine learning element making decisions based on human guidance), and then acting (the process leading to the customer engagement, in the conversion funnel).

“One of the myths with AI is that it is all about fancy mathematical calculations that are baked into the AI process,” commented Steve Casey. “But the truth is that AI is less about the underlying maths, and the various types of algorithms, and much more about the amount of data that is required for AI-powered decisions…..AI excels in locating relationships and patterns in data.”

He added: “Another myth is that AI works right out of the box. In fact, AI solutions need training time to test, learn and optimise….It really is about continuous, rapid iteration. Artificial intelligence will only be able to improve and optimise performance through that process of trial and error.”

AI systems rely on a steady supply of accurate, updated and complete data to be successful.

In a recent global survey of data scientists, when asked about what they spend most of their time doing, 60% said they spend most time ‘cleaning and organising data’ followed by ‘collecting data sets’ 19% and ‘mining data for patterns’ 9%.

Where is the best place to get started with artificial intelligence?

In the short term at least, computers are not yet ready to think in a highly strategic or contextual way. The message that we are hearing from the insurance industry is not to be too worried about all the talk of apps, chatbots or other AI-powered solutions.

Ultimately artificial intelligence will allow us to be more strategic in the goals we can achieve in the insurance business. It’s certainly going to raise the horizon of what can be done with data and automation. By reducing the strain of repetitive and manual work that humans do, it will allow us to increase the time spent on higher-order tasks.

Tough questions, skills and cultural challenges with AI

There is a need to be ready and ask tough questions about machine learning and AI-derived solutions. How does the solution work? What outcomes can we expect? At the business level, all of this is possible without understanding how the different algorithms actually work.

As with any hot technology, there is a need to make sure the insurance business is not just ‘washing’ an already superficial solution with AI. Prepare for the intensive data and customer content required for the test-and-learn cycles.

Because AI is such a voracious consumer of data, there’s a need to make it a cross-department effort. Artificial intelligence involves data and analytics from across the whole enterprise, so everyone – including marketing, the analytics and customer insight teams – will need to be involved. Then there’s a need to trust the answers coming back from the machines. In this way, the insurance organisation will begin to really reap the benefits from AI, over repeated iterations of data. It is not going to be a ‘set and forget’ process.

Follow the link to the LexisNexis Risk Solutions website to find out more about how we support insurance providers.