For insurance brokers, machine learning starts with people

Automation - it’s a word that’s becoming increasingly ubiquitous in the insurance world. Ever since digital transformation gained momentum as one of the 21st century’s most important business trends so far, companies have pursued technology capable of lowering costs and adding value without sacrificing service. Machine learning (ML), a branch of artificial intelligence (AI) predicated on systems’ ability to analyse data, identify patterns, and make decisions, is a key automation tool in this quest. 

Insurance, already an information-dense sector, is experiencing exponential growth in data volume; there is now far more data per organisation than any human could possibly analyse and extract value from, and therein lies ML’s potential. But how is one particular area of insurance - brokerage - benefitting from the application of automation tech? We spoke with Daniel Prince, CEO of Rethink Underwriting Limited (part of Howden Group), and Will Panchaud, Chief Product Officer at Concirrus, to find out more.

What are the benefits of implementing ML?

Insurance has historically been a relatively conservative industry, reluctant to change its centuries’ old infrastructure. The COVID-19 pandemic made it clear that embracing digital technology is no longer an option but a necessity. As such, Prince considers the “journey” of brokers applying ML to have only just begun. “Really simple things like replacing simple, repeatable tasks and filing documents are the current use cases. However, the future should be focused on improving customer service and matching the actions we’re seeing from Big Tech brands.”

Several areas of finance are currently under threat of disruption from tech companies and their associated partner ecosystems, which are able to achieve equivalent levels of service with a more intuitive sense of modern customers’ needs and wants in the digital space. Therefore, the argument for brokers integrating ML must accommodate the aims of both businesses and their clients. Panchaud is convinced these criteria can be met:

Brokers gain

• Speed of response: “ML on the frontline can dramatically improve a broker’s ability to win business. Building and adapting models to support placement decision making can help target the right markets with the right placement structures and therefore accelerating speed of response.”
• Customer understanding: “ML is essential to better understanding the relationship between behavioral data and risk.”
• Product innovation: “Automation facilitates the identification of new product opportunities based on data-driven insights, while also suggesting improvements to current products. Combining this with the Internet of Things (IoT) data can allow brokers to create more dynamic insurance products.”
• Risk advisory and loss prevention: “The risk landscape is rapidly changing due to emerging risks such as climate change and the global pandemic. With access to market data, direct customer relationships, and advisory experience, brokers can provide enhanced risk advisory and loss prevention advice.”

Customers gain

• Tailored products: “Driven by the increased availability of data and modeling, customers can gain access to products and insurance coverage that better suits their needs. This can result in more applicable coverage, fairer premiums, and usage-based policies.”
• Market insights: “Customers can secure market insights based on a collective understanding and market experience, and then benchmark against their peers.”
• Operational efficiency: “Although insurance is a critical protection measure for commercial customers, it’s ancillary to their core business operations. With the evolution of the insurance ecosystem providing digital purchasing, claims, and risk management experiences for customers, operational efficiency gains can be made, not only providing financial protection but also value-added services.”

Priority: Optimise data collection

The business case for ML is clear, but what about its implementation? A solid data collection strategy is both the foundation and engine of automation tech and once again, Prince believes that brokers have barely scratched the surface of its potential. This time, however, few contemporary companies are failing to appreciate data’s importance and are beginning to invest accordingly. “Howden and Rethink started to see the benefits of data collection several years ago, but even then, we still have a long way to go,” he says. Opting to focus on capturing as much structured data about clients as possible, Howden-Rethink then validates and supplements this with third-party data sources. “There is a balance between using data sources to reduce your customer journey and really understanding your client. We are also using AI to extract legacy data, but this is not easy for a complex, non-homogeneous business.”

Regarding this latter point, Panchaud recommends augmenting ML with other automation technologies - optical character recognition (OCR) and natural language processing (NLP) - to make unstructured data (PDFs, Excel spreadsheets, photos, etc.) manageable and ready for modeling. “It is important not to silo this responsibility to a data team within the organisation, but rather mandate responsibility to individual heads of departments,” he adds. “An improvement in data quality does not start with technology; it starts with people, processes, and culture.” This is an important statement and indicates that tools like digital customer portals and ecosystem platforms are invaluable data capture sources because they incorporate the user’s experience and generate value.

Reshaping the customer relationship

Ultimately, integrating ML into insurance brokerage should be about transforming the customer relationship for the better. Consumer anxiety about data collection is tangibly rising - a survey by Adobe of over 5,000 people found that 67% were concerned about identity theft and 57% about data breaches, with 48% adding that they considered breaches an inevitability. By that same token, 76% were comfortable sharing their data, providing that doing so would directly benefit them. Panchaud states that this is the perspective brokers must emphasise: “ML, when applied to risk analysis, can provide insights into areas which represent higher risk within a customer’s portfolio, allowing loss prevention advice to help both mitigate losses and build customer satisfaction, transparency, and trust.”

It follows, therefore, that empathy with clients is particularly important when introducing any new technology. Everyone will have a different reaction to ML, says Prince, and learning to balance its pros and cons will necessitate a decidedly open and honest approach. “Some will love the speed and increased control, but others will feel that their own personal circumstances and personal relationships built up over the years suddenly mean nothing in the quest for market efficiency. Establishing the hand-offs between machine and human will define the winners and losers in maintaining great client satisfaction.”

How is Concirrus helping clients integrate ML?

Panchaud: We work with our customers to accelerate their digital transformation agenda, with a customer base spanning insurers, MGAs, brokers, and reinsurers. Concirrus is invested in realising the potential that AI and Big Data hold for the entire insurance value chain. 

Our Quest platform includes behavioral risk analytics from billions of rows of IoT data and leverages the latest in AI to deploy predictive pricing models that outperform actuarial models. Automation enables our customers to be more operationally efficient and delivers a better experience to their customers and partners.

How is Rethink-Howden implementing ML?

Prince: Rethink is using ML to develop its tools to assist the rapid development of the algorithmic underwriting market. Howden, as a broker, is focused on listening to what our clients are doing to enhance their business and ensuring any new risks are identified and appropriately covered.