Machine learning will revolutionize the future of insurance

Over-regulation, technological changes, changing customer behavior and competitive pressures are driving the disruption of the traditional insurance business model. How should insurers re-think their strategy for the future in an agile market that demands a customer-centric view?

Liberty Mutual leading the way

The insurance industry is under immense stress with threats to its growth prospects looming large.1Over-regulation, technological changes, changing customer behavior and competitive pressures are driving the disruption of the traditional insurance business model. 2 How should insurers re-think their strategy for the future in an agile market that demands a customer-centric view?

We can agree that the backbone of an insurer is its ability to use predictive models to forecast future events and estimate the impact of those events, especially in pricing, sharing risks and claim loss predication. However, in today’s world, that is just a fraction of what it takes to be a leading insurer. In an industry that has lagged in customer innovation, machine learning can be that edge that separate winners from losers. The impact of machine learning can be particularly ground-breaking in the claims management process, starting from how data is used to yield higher predictive accuracy and to increase customer satisfaction during claims settlement process.3

Among the insurance companies, Liberty Mutual stands out in its application of machine learning to improve customer experience. Liberty’s innovation is spearheaded by Solaria Labs, the company’s in-house technology incubator founded in September 2015. 4

Liberty Mutual has developed two applications that leverage machine learning to improve customer experience: a crash-damage estimator application and an enhanced navigation portal that chooses your driving route based on where you are least likely to have an accident.5As an example, let say you have one of those bad days and you a reversing in an unfamiliar parking lot and in your momentarily loss of focus, you backup and hit a light pole. Suddenly you are alert, you get out of your car to preview the damage and wonder how much this would cost to fix. With Liberty Mutual application, you upload a picture and cloud-based computer analyzes the picture and estimates your repair cost. Using artificial intelligence, this application computes cost based on thousands of car crash photos.6 On the other hand, the navigation portal computes data by aggregating public data on auto theft, parking citations as well as crashes using proprietary insurance knowledge, to advise the safest driving routes and places to park in major US cities.7

Machine learning can be further leveraged to address a major cost element for the insurance industry, fraudulent claims. Fraudulent claims cost the industry more than $40billion a year. 8 This can be done by making machines learn from audits of closed claims. Machine learning is capable of computing seemingly unrelated datasets, including structured, semi structured and unstructured data, which would improve the quality of its data used to validate claims and price risks. The traditional process of claim settlement is an internal long convoluted process that requires several touch points by employees.  However, machine learning can give birth to touchless claims that does not require human intervention. This would lead to reduction in claims through earlier identification of the fraud and optimal allocation of resources to manage authentic claims.

Given the dynamic role of machine learning in the insurance industry and the evolution of artificial intelligence at a rapid pace, Liberty Mutual may explore other strategies to remain ahead of the curve. In addition to its inhouse innovation team, Liberty should consider an acquisition strategy of technology companies that have solutions or strategies that aligns with its business model. Start-ups and tech innovators realize the potential applications for machine learning in the insurance industry, which has resulted in innovations that could optimize the mundane processes of insurance companies.9

The advantage of an acquisition strategy is that it allows Liberty to focus on what it does best, which is underwriting and investment activities. Also, the cultural and reward systems of a typical startup or tech innovator, which drives innovation, may not fits Liberty’s existing business model.  Although, the acquisition strategy has its merits, there are concerns that needs to be addressed. This may be an expensive strategy to execute, given inflated valuations of technology companies. Also, the regulatory environment is a key considerations as disruptions from technology companies may not align with the steep requirements of the regulator.

As Liberty Mutual refines its strategy and reflects on the role of machine learning in shaping the future of insurance, it need to determine whether an in-house or an acquisition strategy would give it an edge. (721 words)

 

Endnotes

1PWC, “ Top insurance industry issues in 2018”, February 2018, https://www.pwc.com/us/en/industries/insurance/library/top-issues.html

2PWC, “ Top insurance industry issues in 2018”, February 2018, https://www.pwc.com/us/en/industries/insurance/library/top-issues.html

3Forbes, “ How AI and Machine Learning are used to transform the insurance industry”, October 24, 2017, https://www.forbes.com/sites/bernardmarr/2017/10/24/how-ai-and-machine-learning-are-used-to-transform-the-insurance-industry/

4Liberty Mutual Insurance, “ Apps That Lessen Worry: Liberty Mutual Insurance’s Solaria Labs Unveils New Developer Portal to Benefit Auto Owners”, January 04, 2017, https://www.prnewswire.com/news-releases/apps-that-lessen-worry-liberty-mutual-insurances-solaria-labs-unveils-new-developer-portal-to-benefit-auto-owners-300384202.html

5Liberty Mutual Insurance, “ Apps That Lessen Worry: Liberty Mutual Insurance’s Solaria Labs Unveils New Developer Portal to Benefit Auto Owners”, January 04, 2017, https://www.prnewswire.com/news-releases/apps-that-lessen-worry-liberty-mutual-insurances-solaria-labs-unveils-new-developer-portal-to-benefit-auto-owners-300384202.html

6Liberty Mutual Insurance, “ Apps That Lessen Worry: Liberty Mutual Insurance’s Solaria Labs Unveils New Developer Portal to Benefit Auto Owners”, January 04, 2017, https://www.prnewswire.com/news-releases/apps-that-lessen-worry-liberty-mutual-insurances-solaria-labs-unveils-new-developer-portal-to-benefit-auto-owners-300384202.html

7Liberty Mutual Insurance, “ Apps That Lessen Worry: Liberty Mutual Insurance’s Solaria Labs Unveils New Developer Portal to Benefit Auto Owners”, January 04, 2017, https://www.prnewswire.com/news-releases/apps-that-lessen-worry-liberty-mutual-insurances-solaria-labs-unveils-new-developer-portal-to-benefit-auto-owners-300384202.html

8Forbes, “ How AI and Machine Learning are used to transform the insurance industry”, October 24, 2017, https://www.forbes.com/sites/bernardmarr/2017/10/24/how-ai-and-machine-learning-are-used-to-transform-the-insurance-industry/

Pictures, https://supplychainbeyond.com/wp-content/uploads/2017/09/what-is-machine-learning.jpg

 

 

 

Previous:

Using Machine Learning to Improve Neonatal Patient Outcomes at Cohen’s Children Medical Center

Next:

Steam-powered ideas: a market for open innovation

4 thoughts on “Machine learning will revolutionize the future of insurance

  1. Great read!

    Acquisition or in-house is a choice that often troubles companies when developing new capabilities. Because leaders in the industry are usually not part of the company, I often lean to acquire knowledge rather than growing. I think this is particularly necessary in a rapidly changing industry such as machine learning, where the time to develop new in-house knowledge might have made the said in-house developments redundant.

    Thank you Ulunma for a very interesting read.

  2. Very interesting article! I particularly like how Liberty mutual is utilizing machine learning to improve the customer experience through their crash-damage estimating capability. When I think of insurance and machine learning, I typically think of complex algorithms that will try to determine exactly how likely a person is to be a liability above and beyond their premium. Their use of machine learning is something so customer facing reveals how much a firm can miss out on if they only focus about machine learning in the context of their core capability. Here they are using ML to generate a new, customer friendly capability that would have been impossible before!

    Thanks for the great read!

  3. Given that the algorithms used to address Liberty Mutual’s challenges need to be built on Liberty’s data and will be key to the firm’s ability to compete long term, I believe that the capabilities need to be built in house rather than through acquiring a company that is not tailored to meet Liberty’s needs. These capabilities are built by acquiring and retaining top data science talent. Additionally, I think machine learning can best be leveraged by Liberty to enhance their underwriting rather than to simplify claims. However, the main question with regard to machine learning in this space is whether it is a sustainable source of competitive advantage as large insurers gather more and more data and models become more of a commodity.

  4. While solutions like crash-damage estimators can be a good way to smooth out the initial customer engagement process, I think there is also danger in having it provide a potentially unrealistic quote, and sabotaging the customer experience from the start. For an event like a car accident, which could have drastic financial consequences for the average, unprepared American1, the insurance companies need to take every precaution to ensure that the customer does not receive unrealistic financial guidance. While I agree that machine leaning can be a useful tool in helping users better understand the context of their accident (i.e. which parts are more critically damaged, and might take more time/money to repair), I would veer away from providing any numerical suggestions. In the worst case scenario, if the software underestimates the true cost, then the app will be pitted against the insurance company and/or the repair shop. While we could note that the app only provides estimates, it could still leave a bad taste for customers who view it as an official extension of the insurance process.

    I find the enhanced navigation solution to be particularly intriguing, although this is a feature that needs to be extensively tested before a direct beneficial relationship can be established. For example, even though the optimized route may be objectively safer, would using it lead to a higher incidence rate if the driver is unfamiliar with these roads? Alternatively, if a user takes the optimized route and gets into an accident, then would the insurance company bear some of the blame? Presently, most machines cannot solve problems without proper human context, and in a scalable business like insurance, where machine learning could definitely offer valuable insights over large datasets, I still think there is a while to go before these solutions can be made customer facing.

    1. Nearly 60% of Americans Can’t Afford Common Unexpected Expenses, Bankrate. (2017, January 12). Retrieved from https://www.bankrate.com/pdfs/pr/20170112-January-Money-Pulse.pdf

Leave a comment