Written by: Trevor Lloyd-Jones, Content Manager, LexisNexis Risk Solutions

The majority of artificial intelligence applications require vast volumes of data, both in the form of training data (to prove an algorithmic model) and then live data, which must be gathered, stored securely, normalized, annotated, analysed, and presented to end-users in a meaningful way. In order to take advantage of what artificial intelligence has to offer, it’s going to take insurance providers overhauling their technical foundations, workflows, and strategic goals.

There was agreement at the recent Insurance Analytics and AI Europe conference in London that there is now a real push to realise more analytical value from current and historic insurance data.

For artificial intelligence to work, the quality of the data is all important. Even if we have the right infrastructure, for the insurance process to work, it must have the right data, directed to solving the right question for the business.

There are many more live projects in advanced analytics – pre-algorithm digital testing, to proof-of-concept, to private beta and live services – compared to a year ago, and a more pragmatic approach to test and build.

As Craig Beattie, Senior Analyst at Celent, expressed it in his opening presentation at the conference: “There’s a sense that AI is the part of the technology that doesn’t work. If it’s working, it seems to suddenly become part of technology….In terms of where the opportunity is currently being realised, claims management, marketing, product development, sales and distribution are the top areas ranked as most important for AI by insurers.

“Data science is especially popular for front-end [customer-facing] functions in insurance and to manage risk. But there’s been a significant shift in the conversation around insurance and AI this year. It’s becoming very, very pragmatic. Outside of fraud and claims optimisation, there’s now a long list of issues we can solve with AI, and real projects.”

Insurers becoming more pragmatic with AI, testing more business processes

There’s certainly a lot of rhetoric in the media and eye-catching headlines about the way algorithms are going to change our lives, and the challenges of AI. Just recently, the UK Competition and Markets Authority (CMA) published an economic working paper on the use of algorithms by insurance and other businesses operating on online markets and how they could be used to achieve a tacitly coordinated pricing outcome, potentially in breach of competition law.

In its investigation, the CMA found what is said was widespread use of algorithms to set prices, particularly on online platforms, where “many sellers” use them. As well as the simple pricing rules provided by the platforms themselves, the CMA discovered that some third-party firms sell more sophisticated algorithms to retailers or take on the role of price determination using computer models on behalf of their clients.

The feeling in insurance is that there is a path to be found where AI can produce many statistical models and experimentations with very real consumer benefits.

Starting with natural language processing (NLP), bots and virtual assistants, which are really the foundations of insurance automation rather than AI itself, we are starting a new journey. How many actuaries are changing their job titles to data scientists? How many underwriters are moving to head up dedicated innovation and AI teams?

There was a discussion at the Insurance Analytics and AI Europe conference about the current barriers to implementing artificial intelligence and where AI should sit within the insurance organisation. A poll of the attendees showed a cross section of their views around the challenges and opportunities for AI in insurance.

Support is growing for AI processes to sit within a dedicated AI or innovation team within insurance – as a change agent, internal consultancy or task force, piloting, scouting the business and testing – rather than in fraud or in the IT department as would have been the case a couple of years ago.

Where should AI sit within the insurance organisation?

  • Under the Chief Data Officer (CDO) 30%
  • In the AI team 27%
  • In the innovation team 15%
  • Under the Chief Operations Officer (COO) 15%
  • Other 13%

Source: Insurance Analytics and AI Europe

What are the barriers to implementing AI in the insurance organisation?

  • Getting AI into production and into live services 31%
  • Skills and talent 22%
  • Legacy systems 20%
  • Data 9%
  • Getting past the IT department! 9%
  • Other 9%

Source: Insurance Analytics and AI Europe

How does the prospect of AI make you feel?

  • Excited 51%
  • Cautiously optimistic 34%
  • It’s the end of the world! 7%
  • Neutral 4%
  • Concerned 4%

Source: Insurance Analytics and AI Europe

Insurance is fundamentally based on data processing and predictions. In the final analysis, considering insurance and beyond, AI is going to make us all more human.

Today people are doing many repetitive tasks where ultimately we will not be able to compete with machines, especially tasks requiring the highest volumes of data input or seeking patterns and inferences in complex problems. It is going to be in the insight, investigating the business context of the data and in the human touch, the customer contact that humans are going to excel in insurance.

Increasingly automation, and then AI at the higher level of decision-making, is touching all parts of the insurance value chain. There’s a lot of talk about it being the future of pricing and underwriting, though first, of course, must come the strong foundation of tried and tested data, clean data in the ecosystem.

AI is able, theoretically at least, to deliver a better prediction rate, especially with higher data volumes. Machines are better at finding patterns in data, and the more data, the sharper will be the pattern recognition.

Machine learning methods are ultimately going to be better at modelling deep interactions in data, and they’re going to be faster at assimilating information embedded in a customer interaction, and serving it on the table to the underwriter.

With AI we are moving slowly towards near real-time pricing models and a deeper, ongoing relationship with the policyholder, beyond the annual policy renewal. This is already evident if we think about motor telematics and pricing in the context of real driver behaviour, based on the continuous inflow of data. In claims we can think of automated photo recognition for motor repair estimation, or sophisticated new data sets and alerts to emergency services going into FNOL (first notification of loss) processes.

All of this is technically possible although most business models are not yet ready for it. Within the next few years we are likely to see real customer behaviour, customer attributes and new data sets, influencing underwriting processes and the insurance business model, as we are seeing in motor telematics. Right now, machine learning and AI models still do not outperform human-based underwriting.

As Anne-Sophie Grouchka, Chief Customer Officer at Allianz France, commented in the meeting: “AI cannot be a self-sufficient technology. It must be part of a global journey and we need to build that journey. For any insurer AI will be the heart, the core in future. It’s the question that will reshape the insurance business in the future….When it comes to pure customer services AI offers a lot of opportunities to support enquiries and speed up processes.”

But she added that there are potential conflicts between machine-defined processes, AI, and human processes. AI will not necessarily replace humans in the customer service setting, rather it will help customer service issues for example in stratifying large volumes of emails or other information, making humans more efficient, doing what they do best.

“The quality of data is very important….As an industry we know how underwriting and claims work. But we can’t yet say we know how AI works,” commented Somesh Chandra, Chief Health Officer & board member MAXIS & AXA EB Partners.

Insurance customers are just waiting for a simple answer to their product query or to their claim. They don’t need to know about all the sophistication beneath the surface. But it’s our job in insurance to join this revolution and make it a reality. It’s the machine, together with the human handler, that is going to drive the best performance.

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

Post a Comment

Your email address will not be published. Required fields are marked *