Artificial intelligence and insurance: Part 1
Insurers are accelerating the integration of artificial intelligence solutions across their operations, but how are successful early adopters deploying the technology and where are they headed?
A tech revolution is gathering pace across the insurance industry; one that could radically transform the risk transfer value chain. Artificial Intelligence (AI) technologies that enable computer systems to perform tasks normally requiring human intelligence, such as visual perception and decision-making, are being plugged in right across the sector.
Machine learning (ML), a branch of AI that provides computer systems with the ability to learn and adapt independently, based on algorithms and the analysis of data, is being deployed in operations from marketing to claims handling.
Gurpreet Johal, global reinsurance and London market leader at consultancy firm Deloitte, says that AI integration is variable across the insurance industry.
“In most areas of underwriting and in pricing models, techniques haven’t changed that much despite the advancements in technology and the large amount of data that is available to insurers,” he says.
“But some pockets of the industry, for example in the London market with the emergence of ‘algo syndicates’, or in certain regions, they are at par or are quick followers compared with hedge funds, for example, or retail banks in North America and Asia.”
“If these [insurance] companies scale up they could have the same impact on the sector that the quant funds had in the ‘80s and ‘90s, for example,” Johal believes.
Alan Tua, portfolio director at Ki Insurance, Lloyd’s first fully digital and algorithmically-driven syndicate, agrees: “While a lot of the well-trodden approaches in the actuarial world are standard in the introductory chapters of ‘Machine learning 101’ textbooks, where the insurance industry does lag is in getting onboard with some of the technology that allows business to turbo-charge these approaches and embed them into end-to-end production systems that operate in the real world.”
But early adopters of AI and machine intelligence (MI) are already seeing benefits in select areas, such as faster claims settlement, more targeted cross- and up-selling, improved fraud detection and better risk scoring, according to Swiss Re Institute’s head of research and data support, Jonathan Anchen.
“The insurance industry needs to build on successful pilots to deploy AI in multiple processes across the firm. [But] to reap large scale benefits from AI, not only is it necessary to invest in digitising firm-wide operations but also to break down data silos and dedicate resources to integrating models and algorithms into workflows,” he says.
Pricing, underwriting and portfolio management can benefit from intelligent intervention through analytics and robotics, says David Ovenden, Willis Towers Watson’s global director of pricing, product, claims and underwriting.
“It means reducing unnecessary human tasks like re-keying, and providing contextual underwriting decision support, for example,” he explains.
“If ML or analytics is added into a process it becomes possible to make intelligent interventions; routing claims, for example, to the most appropriate team or person. Adding rich analytics to rule-based judgements can improve operational decisions, identifying when a survey is needed.”
Another area where MI is being applied is in agent recruitment and retention, according to Swiss Re Institute’s Anchen.
“Insurers initially ‘M-enable’ systems to identify individuals most likely to become successful producers. These systems can also improve producer-client matching. For example, real-time, automatic matching of call centre agents to members with whom they are likely to have the highest affinity.”
Smarter mechanisms for underwriting triage and routing can be more effective than current business rules in the life and health business, Anchen says.
“For example, triage between depths of investigation, i.e. full versus simplified underwriting, safely waive additional evidence, such as lab tests or allocating referrals to the right place in the organisation (e.g. junior underwriter versus medical officer).”
Use cases keep cropping up. The Spanish motor insurer Admiral Seguros already uses an AI tool to create damage valuations and generate an offer of immediate payment on damaged vehicles. Developed by the tech company Tractable, the tool evaluates vehicle damage with photos sent through a web application. The app, via the AI, completes the complex manual tasks that an advisor would normally perform and produces a damage assessment in seconds, often without the need for further review.
AI has featured in several InsurTech start-up models. In the US, Atlanta-based layr.com, which participated in the first cohort of the Lloyd’s Lab initiative, uses its proprietary price and appetite prediction engine to match businesses with policies from carriers at the different prices. Unlike comparison engines or aggregators, prospects can discover what coverage they need and also what coverage is purchased by similar businesses.
Specialty carrier Beazley recently enhanced its reputational risk insurance policy by providing an AI tool that helps policyholders control their corporate brand and reputation more effectively, benchmarking against competitors and peers. The cover, underwritten by a consortium at Lloyd’s, includes crisis management services.
Sound and vision
It’s likely that insurance industry players are only seeing the tip of the AI iceberg, according to Joan Cuscó, global head of transformation at Madrid-based global insurer MAPFRE.
“The current maturity of deep learning provides high-level accuracy in models that are trained using images. Picture and video are critical for both underwriting and claims, so we’re seeing a tech race in the insurtech scene around image processing, including aerial and satellite imagery,” he says.
Natural language understanding (NLU) is another area of evolution to watch.
“Artificial intelligence, driven by machine learning and its many variants, has made real gains in providing predictive insights from structured data, which is generically numerical data. But 80-90% of the data that insurers rely on is unstructured, expressed in the form of language data,” explains Pamela Negosanti, head of sales and sector strategy (FSI) at the tech firm expert.ai.
“The ability to scale data-driven automation and industry expertise will be the main watchword in the challenging times ahead. Advanced NLU capabilities will continue to demonstrate measurable benefits within 2021 from loss avoidance and unintended risk exposure reduction to operational efficiency and processing time reduction, with consequent drastic savings,” Negosanti says.
Deloitte’s Johal thinks that the full spectrum of risk transfer will be changed by AI technology, from alternative reinsurance capital through to small business product design.
“Risk trading and ILS continue to attract private equity houses, hedge funds and other capital. They’re very sophisticated and open to AI integration,” he says.
“Alternatively, in the SME insurance market there is the challenge of acquiring customers and servicing them effectively, designing new products such as parametric insurance. AI can play a part there. This has the potential to disrupt some traditional products - and [their] carriers.”
In Part 2 of this article series we will look at the challenges confronting insurers who want to scale up AI integration, and how to tackle them.