Author: Freddie Scarratt
Many topics drove the InsurTech narrative in 2023, but perhaps the one that grabbed the most universal attention was artificial intelligence (AI). The graph shows the interest in AI over time, as tracked by Google (i.e., showing interest across all industries).1 The growth in the past 12 months has been exponential, with popularity peaking in late 2023. This peak coincides with the infamous OpenAI leadership transition and the swift return of the CEO in November last year.2
AI is not a new subject — the term was first coined in in 1956 at the Dartmouth Conference, where researchers discussed the possibility of creating machines that could exhibit human-like intelligence. Early developments in AI can be seen in the 1940s and 1950s with the work of pioneers such as Alan Turing, who proposed the idea of a "universal machine" capable of simulating any other machine's behaviour. AI research progressed in the following decades with the development of various techniques and algorithms, including symbolic reasoning, expert systems, machine learning and neural networks.
While AI has experienced periods of both excitement and disillusionment (the latter known as AI winters), it's seen significant advancements in recent years. These advancements have been consequential enough to arouse the interest of regulators — the UK's National Cyber Security Centre published Guidelines for Secure AI System Development in late 2023.3 Large datasets, large language models (LLMs), increased computing power and advancements in deep learning techniques have contributed to the rapid progress of AI in various domains. Furthermore, and most importantly, it has acquired a 'brand' — and is part of everyday language for the first time in its history.
Today, AI is a rapidly evolving field with applications in numerous industries. The best way to judge its likely impact upon (re)insurance — and within InsurTech more specifically — is to look at its practical applications within our value chain. How might this technology affect our distribution, pricing/underwriting, business operations and the claims process?
Distribution
Can the introduction of AI boost growth by increasing distribution? In short, we think the answer is 'yes'. One area of particular interest is data-driven personalisation as consumers become increasingly reluctant to spend. AI can determine the optimal customer offer and select the most suitable distribution channel via automated workflow tools. This personalisation maximises customer acceptance, leading to increased revenue and market share for insurers as insurance intermediaries and carriers have more time for customer service. Additionally, AI enables insurers to identify the most efficient and cost-effective distribution channels, resulting in reduced expenses.
Pricing/underwriting
Pricing is central in all (re)insurance decision-making, whether it be the consumer, small to medium-size enterprise (SME) business owners or large multi-national insurance carriers purchasing their reinsurance. Our industry is based on having a combined ratio less than 100%, therefore successful innovation in underwriting and pricing to better price is paramount.
The hope is that sophisticated proprietary platforms can bring customers and insurers closer together, providing customers with unique experiences and customisation, and a premium that aligns with their level of risk. AI offers the prospect of determining pricing in real-time, based on usage data and a comprehensive evaluation of risk, with the system able to access risk factors and customer information from a range of sources without lengthy submissions from the customer and their broker. This approach will also enable carriers to effectively manage their exposures, correlations and potential maximum losses (PMLs), resulting in a more comprehensive and efficient capital management process.
One provider offering such AI-driven services is SYPHER, which operates in the coastal Property insurance market. Its offering includes risk transfer products and processes, with streamlined approaches to structuring, accounting, claims, analytics and sidecar transactions.
Business operations
AI-powered InsurTechs such as SEND, ProNavigator and Federato are helping to establish AI within the operational platform of InsurTechs, new market entrants and the incumbents within the (re)insurance space. Their vision is to create '360° platforms' that can be used to manage multiple processes relevant to a business. SEND's Underwriting Workbench is a great example — a single platform for managing new business, renewals and endorsements. The companies using such platforms (Everest, Renaissance Re, Convex etc.) hope to more efficiently track workflows, eliminate 'rekeying' and overall, speed up the underwriting process.
Claims
One area in which we have a keen interest is AI's use within the claims workstream. This area is of particular importance to clients, and AI offers the potential for significant efficiency savings. It also has multiple use-cases — whether in customer experience, fraud detection or workflow management. AI-powered customer chatbots have been the torchbearer for this in recent years, however companies such as Shift Technologies and Sprout.ai are helping (re)insurers identify potential claims fraud and providing detailed contextual guidance to empower investigators, while offering a seamlessly integrated claims experience across multiple channels with AI decision support.
The key to all the above is data. It's the foundation for AI usage, and 'data driven insights' is one of those buzz phrases most seen in articles on harnessing AI. Of course, the importance of data is nothing new to (re)insurers, who have collected, stored and drawn insights from it for decades across underwriting, claims and pricing.
But this collected and stored data creates a problem. The data stored in the primary legacy systems of most insurers is not yet suitable for AI applications. These systems are often isolated and have inconsistent definitions of data types, making it challenging to use them effectively. Adding to the challenge is the quality of the data, which can vary significantly. It's within this interplay between the construction of the data and where it is held that the challenge lies. AI must be able to identify, access and use data in order to add value.
Our view is that, given these challenges with data and legacy systems, AI's practical application as a game-changing technology across the (re)insurance value chain may be overhyped at this stage. After all, the internet didn't dramatically change economic productivity, even if it did change behaviour and business practice. The key is in clearly and concisely articulating its value for each part of the value chain. We can see this with some successful examples already, such as Shift Technologies' adoption of AI in fraud detection. We will be keeping a close eye on AI progress this year in our clients' core business practices (and our own).
