AI and insurance: who’s making the decisions?
Brian Mullins is the CEO of Mind Foundry, an AI solutions provider in Oxford, UK. He is an expert on bleeding-edge technologies in the insurance space. We spoke to him about the benefits of AI within insurance, and its role in bringing embedded products and services to market.
How embedded is AI in the current insurance market?
AI is very embedded in the insurance market, especially compared to other sectors within financial services. Several breakthroughs have been made largely by disruptors, who have used AI to change insurance methodologies and that investment continues to optimise their business practices, widen their competitive edge and offer customer-centric products.
However, we will look back ten years from now and see that we were only beginning to scratch the surface of what is really possible. The desire to change and implement all of the benefits of AI are tampered with the reality that many insurers are large organisations with workforces and work practices that will take years to adjust. This is why much of the innovation to date has been the preserve of insurtech disruptors.
Why has AI become so central to the underwriting/customer services process?
Current data volumes mean that processing quotes, renewals, claims, customer requests, etc. is at the limits of what can be handled by human operators assisted by traditional information systems. In the past decade, companies have turned to robotic process automation to speed up manual processes and amplify the work of their handlers. Yet today this is still not sufficient, and to survive in these oceans of data, let alone to innovate and stay ahead of the curve, insurers like many other businesses have turned to AI to discover more subtle patterns in data, automatically adjusting to new trends, threats, and opportunities. Machine Learning algorithms can be trained to discover policyholders’ behaviour, preference, inclination for risk, and even fraud, learning from past examples and adapting to new ones, without the need for human intervention.
What trends in technology are we seeing emerge?
Usage-based insurance has been a central part of insurance portfolios for a while now, especially in car insurance, and at the heart of this is the accurate assessment of individual risk. This has paved the way for a rise in driving-behaviour classification models, driven by telematics data, that capture the capabilities of connected vehicles and edge computing. Access to real-time data is enabling insurers to reimagine their insurance products and deliver hyper-personalised insurance offerings.
Another emerging trend – which we have seen in other industries too – is the use of AI to speed up data processing. This reduces the human time needed to process claims or investigate fraud, which allows insurers to concentrate their human resources on validation, correction and non-trivial investigation.
However, it is worth noting that some of the AI trends that were starting to emerge – like computer vision – could get stifled by much-needed regulation, such as the recently published Proposal for a regulation laying down harmonised rules on AI by the EU. Examples of this include Lemonade’s ML model which uses facial recognition to assess if a claimant is trustworthy, or applications that include the use of computer imagery to instantly assess damage to a car or help advise a customer on the medical services to seek based on their symptoms.
Hyper Automation is emerging too. What differentiates it from ML, RPA and AI and why is it becoming popular?
Automation in general has been used extensively in insurance, and the use of Robotic Process Automation (RPA) is common, if not assumed. RPA automates specific tasks by using a set of instructions. It is particularly efficient when it comes to engaging with customers as the use of chatbots and auto-generation of claims reposts free up customer service agents’ time to handle more complex cases.
It can’t, however, evolve or learn from new information in the way that AI can. Hyper automation’s end-to-end efficiencies are what have made it so popular. It can be considered an extension of RPA which pulls AI and ML technologies, process mining and decision-making to produce end-to-end automation solutions.
