Global Business Intelligence

Produced by:


Machine learning and insurance: the emergence of a pricing revolution

Artificial are a London-based company using AI to take insurers on the journey to frictionless, digital pricing.


A combination of cumbersome paper-based processes, a lack of transparency and an undeniably poor customer experience is how most would describe the purchase of commercial insurance products today. The team at Artificial are looking to change that, with their development of Machine Learning pricing being just one example of disruptive technology that is delivering better outcomes for customers.


Matrix pricing and the risk of high exposure


Traditional commercial insurance pricing models are simplistic and make use of basic matrix systems which use a small number of variables to alter the premium.


Matrix pricing models often lead to uncontrollably high exposure to risk and large loss ratios, with no guarantee that a given policy price is indeed linked to the actual risk. The lack of additional data inputs, as well as a strict set of constraints, means that insurers miss out on a level of pricing accuracy that could be achieved with data-driven assistance.


Another cause of high loss ratios is the use of pricing bands. If a pricing band captures a large range of customers, they will all be offered cover at the same price regardless of whether their risk places them at the top or bottom of the risk bracket. Customers which represent a disproportionately high risk can therefore find themselves in a lower pricing band, leaving the insurer with loss ratios that are substantially higher than they should be.


The problem with ML pricing


With the latest developments in artificial intelligence and Machine Learning, insurers can begin to optimise their pricing processes and allow underwriters to concentrate on higher value analysis.


Using these techniques it is also possible to cut out vast swathes of expensive processes whilst still adding value. To streamline the customer experience and avoid repetitive back-and-forth questioning, insurance companies can draw on a wealth of third-party data to supplement their algorithms.


Unfortunately, ML pricing doesn’t come without its challenges – many of which Artificial have outlined in ‘Part One: How can I use AI to improve my company’s pricing?’. For insurance companies that have an existing commercialised product on the market – with its own established pricing structure – ML can create more problems than it solves.


Making the switch to ML blindly can create a ‘black box pricing’ model, where algorithms are trained using vast amounts of data and variables, making the calculations less transparent to the insurer and ultimately less explainable to the customer. For example, a price generated by twenty different risk factors becomes very difficult to explain to a consumer who wants to know why they are suddenly paying more than their neighbour for the same product.


Besides, new pricing engines often conflict with the existing pricing structures insurers already have in place. If a pricing model has been established for a product that’s already on the market, it’s hugely inconvenient – and costly – to throw it out and start again.


Making the most of data at Artificial


At Artificial, the team of data scientists are applying ML techniques to help take insurance companies on a journey to full digital pricing and negotiation. The company has developed systems to implement ML pricing models for both new and existing commercial and consumer products without the chaos of a restructure – showing that, in the right hands, ML can be hugely beneficial to the insurance industry. To develop a new pricing model, Artificial makes use of internal and third-party claims and policy data to train the algorithms.


With access to new streams of data, the team are beginning to generate insights into the causes of claims and unexplored risk factors. Johnny Bridges, Artificial co-founder and CPO, explains their recent analysis of telematics data:

The data science team recently used ML techniques to analyse telematics data that showed, in the case of motor insurance, one of the largest causes of claims was due to a significant change in driver altitude (so for example, driving up a steep mountain would increase the risk of damage). When you combine such insights with other sources such as weather data, one can begin to paint a much clearer picture of the underlying risk.
Johnny Bridges, Artificial co-founder and CPO

As insightful as this is, the process inherently brings other difficulties. The entire operation of capturing the wealth information required and converting it into a format that is consumable by a pricing model can be time-consuming.


To streamline this process for insurers and automate the submission intake process of unstructured submissions, Artificial built a Submission Tool. This Machine Learning model uses Natural Language Processing, information extraction, text classification, and named entity recognition to read structured and unstructured documents and extract the data required for a typical insurance policy submission (company names, addresses, policy limit, SOV and Loss Runs).


It also focuses on automation of structured, semi-structured and unstructured documents, allowing insurers to extract key information to ultimately give a quote on demand, without the need to type in any information. The output of all of the information ingested can be returned as structured data consumed by other applications from the Artificial suite or as APIs.


Integration and retaining existing structures


As exciting as ML-powered pricing sounds, applying this new technology often requires significant upfront investment and implementation can drain time and resources. To secure the benefits of smart pricing analysis with minimal disruption, Artificial integrates its pricing engine with an insurer’s legacy systems – saving time, effort and money.


For example, if premiums for an insurer’s existing products are broken down into discrete pricing bands, Artificial’s pricing engine can assign future premiums within the same parameters of those bands – this keeps the original systems and pricing structures in place.


The pricing engine then performs a price allocation, ensuring risk sensitivity and the selection of an optimal price for all customers considering their predicted risk. As Artificial explain in ‘Part Two: How can I use AI to improve my company’s pricing?’, this ensures minimal disruption to current processes but is equally a highly tractable system.


It is also vital that insurance companies have the ability to create adjustable parameters within their structures to constrain the amount that ML outputs can deviate from the original price allocation.


Prices for each band can then be moved up or down by a certain percentage to meet loss ratio targets, with limits imposed on the allocation of new price bands. This allows a smooth transition to the new pricing model for the insurer whilst still providing the necessary levers to tweak prices for increased profitability and market competition over time.


By clarifying the variables driving each assigned premium, customers then have a clear explanation of why their products are priced at a certain amount and insurers themselves have greater visibility and the ability to assign appropriate future prices based on key risk variables.


Scale and deliver with Capita


So how does Artificial plan to scale and deliver this ML technology to established insurers? One of the key challenges start-up companies face in traditional markets – such as insurance – is gaining traction with large incumbents.


Artificial believe the solution is not to compete with incumbents in the insurance market, but rather to work with them, equipping them with the digital tools needed to meet evolving customer expectations and remain profitable in an increasingly challenging sector.


By partnering with Capita Scaling Partner, the corporate venturing arm of Capita Plc, Artificial can make use of a dedicated business development team and benefit from exposure to Capita’s extensive list of insurance clients. Damian Arnold, CEO of Artificial, explains the benefits of such a relationship:

Since we signed the partnership, our innovative solutions have strengthened Capita’s existing insurance offering, opening up new opportunities for Capita’s clients to increase insurance distribution, process business more efficiently and try out innovative new tools. Capita have been an invaluable resource and continue to help us grow.

Damian Arnold, CEO of Artificial

The future of insurance


What, then, does Machine Learning mean for the future of insurance? It’s not inconceivable that customer behaviour will be influenced by changes in pricing. Parametric insurance (or ‘on demand’ insurance) could have a dramatic impact: it may generate scenarios where environmental factors deter motor customers from driving in a particularly risky location or in bad weather because their on-demand premium could rise as a result.


Overall however, the changes will be positive. Customers should receive a more accurate price, and insurers should be able to manage their risk portfolios more accurately and transparently. There is no way of knowing how customers will perceive this new approach. How to determine the impact on profitability and the respective price elasticities of a product’s offering, particularly across an insurer’s entire portfolio, is perhaps the most important question for insurers. Artificial is developing a model to assess and predict this – but that’s an article for another day. Meanwhile, it’s clear that data, AI and Machine Learning are vital to the future of insurance – and that those who understand how to leverage this technology will have the most success.


For more information about the benefits Artificial’s toolkit could deliver for your business, please contact Damian Arnold on