Machine Learning: Powering the Insurance Industry’s “Buy It Now” Button

The rise of big data and machine learning has created new opportunities for incumbent insurers and competitors alike.

Market Dynamics: Insuretech Competitors Emerging

Sales of individual life-insurance policies in the United States have declined more than 40% since the 1980’s [1]. Meanwhile, the way one purchases life insurance has remained largely unchanged for decades. The traditional purchasing process is long and cumbersome, often including multiple interviews, lengthy paper applications and a medical exam. In all, it can take months for an applicant to be approved.

Equipped with an increasing amount of digital data sources and supported by advances in machine learning technology, a number of up-start “insuretech” companies (e.g., Bestow, Ladder, Ethos [2] in life and Lemonade, Slice Labs and Cover in the property and casualty market) are investing in algorithmic underwriting capabilities to support direct-to-consumer business models. In addition to eliminating friction in the purchasing process, start-ups hope to reduce operating expenses (i.e., selling and underwriting costs) through the implementation of artificial intelligence and machine learning.

As a result of this perceived market opportunity and advances in technology, venture capital interest in insurance has skyrocketed. Per Willis Tower Watson’s industry report, venture investments in insurance startups grew by more than 7x from 2013 to 2017 while the number of venture capital investors more than doubled. While some of these dollars are flowing into alternative technologies, a sizable percentage are related to artificial intelligence and machine learning.

MassMutual’s Response

To address these market dynamics in the short term, MassMutual, a Springfield-based life insurance mutual company founded in 1851, has established Haven Life (“Haven”). Haven is a direct-to-consumer affiliate company which has built an algorithmic underwriting capability that allows the company to quote and bind (i.e., issue) a policy in a matter of minutes. A Haven Life application has approximately 30 questions which Haven supplements with third-party data from sources such as motor vehicle records, prescription histories and the Medical Information Bureau [4]. In some cases, the model is insufficient and a human underwriter is needed. However, Haven expects human-reviewed applications to drop as a result of continually “reassessing the data and rules the service’s AI relies on”, per Chief Technology Officer, Todd Rogers.

Building on this capability and to position the company for the medium-term, MassMutual has launched LifeScore Labs to distribute MassMutual’s risk models to the market [5]. The models have been trained on a historical database of MassMutual life insurance applicants. In February and March 2018, LifeScore announced distribution partnerships with iPipeline, a provider of software solutions for the insurance and financial services industry, and SwissRe, a global reinsurer which will utilize LifeScore as an analytics option within its own underwriting system, respectively [6/7]. Customers of LifeScore will have access to applicants’ risk score, risk visualizations as well as details on contributing risk factors. If successful, LifeScore Labs will diversify MassMutual’s sources of revenue and deepen LifeScore’s data set which should in turn improve the models used by its customers and MassMutual.

The Future

In addition to underwriting and distribution, machine learning will impact MassMutual’s product innovation, asset management business and claims processes. As a result, expect the Firm to seek investment across the following categories to position itself for the future:

  • Partnerships: MassMutual’s algorithmic underwriting and distribution capability has positioned the Company for unique partnerships. For example, given the growing gig economy, Haven might pursue partnerships with companies like Uber, Lyft, or Thumbtack in which Haven could sell life insurance policies to workers who lack traditional benefits.
  • New data sources and product innovation: MassMutual competitor, John Hancock, announced in 2016 that it would only sell “interactive” policies that track fitness and health data through wearable devices and smart phones. Policyholders receive discounts for exercising and are entitled to other perks by logging their healthy food purchases. In addition to engaging with its customers, by widening its proprietary data set, John Hancock may be able to more accurately price mortality risk in the future [8]. MassMutual may consider a study of similar health data to bolster its underwriting capabilities.
  • Equipping other channels: Due to the complexity of insurance products, agent-driven sales and omni-channel experiences will continue to be the core of MassMutual’s business in the near to medium-term. The Company likely has/may continue to invest in AI-driven customer service tools such as chat bots and/or sales assistants like Tact.ai.
  • Asset management: A critical component of an insurer’s business model is its negative cash conversion cycle which allows it to invest the premium it receives from customers for financial gain. The use of AI/ML in asset management is burgeoning and applications include: compliance (KYC/AML), sentiment analysis (e.g., sentiment from news articles or social media) and technical trading strategies [9]. Oppenheimer Funds and Barings are investment manager affiliates of MassMutual who will continue to be impacted by these advances and will likely face decisions to either build tools internally or seek third-party fintech solutions.

In light of these trends, a couple questions come to mind:

  1. If any, what will be the role of a trusted, human financial advisor in the future?
  2. What will protection (i.e., insurance products and/or annuities) of the future look like?

(word count: 864)

[1] “Totally Changed Industry Landscape Only Three Years Away.” Carrier Management, 1 Sept. 2017.

[2] Marquand, Barbara, et al. “Fast Life Insurance: Where to Find Instant Coverage.” NerdWallet, NerdWallet, 29 Oct. 2018, www.nerdwallet.com/blog/insurance/instant-life-insurance/

[3] “Quarterly InsurTech Briefing.” Willis Tower Watson, May 2018.

[4] Staff, VentureBeat. “How One Company Learned to Reinvent Itself Daily in the AI Age.” VentureBeat, VentureBeat, 6 Oct. 2017, venturebeat.com/2017/10/06/how-one-company-learned-to-reinvent-itself-daily-in-the-ai-age/.

[5] “About Us.” LifeScore360 , MassMutual, www.lifescore360.com/About-Us.

[6] “MassMutual’s LifeScore Labs and Swiss Re Partner to Bring LifeScore360 to Market.” MassMutual.com, MassMutual, 20 Mar. 2018, www.massmutual.com/about-us/news-and-press-releases/press-releases/2018/03/19/14/15/massmutuals-lifescore-labs-and-swiss-re-partner-to-bring-lifescore360-to-market.

[7] “Ipipeline and MassMutual’s LifeScore Labs to Instantly Deliver Risk Scores for Underwriting.” Ipipeline.com, IPipeline, 28, Feb. 2018. https://www.ipipeline.com/insurance-software-solutions/news/ipipeline-and-massmutuals-lifescore-labs-partner-to-instantly-deliver-risk

[8] Barlyn, Suzanne. “Strap on the Fitbit: John Hancock to Sell Only Interactive Life…” Reuters, Thomson Reuters, 19 Sept. 2018, www.reuters.com/article/us-manulife-financi-john-hancock-lifeins/strap-on-the-fitbit-john-hancock-to-sell-only-interactive-life-insurance-idUSKCN1LZ1WL.

[9] Kolanovic, Marko. Informing Investment Decisions Using Machine Learning and Artificial Intelligence. J.P. Morgan, www.jpmorgan.com/global/cib/research/investment-decisions-using-machine-learning-ai.

Previous:

Man or Machine? Does AI have a place in Venture Capital?

Next:

Virgin Hyperloop One: Will Open Innovation Lead to Its Reality?

5 thoughts on “Machine Learning: Powering the Insurance Industry’s “Buy It Now” Button

  1. I want to respond to your question around the future of insurance, given how it will be shaped by technology in the years to come. I think one major shift – in addition to faster/more efficient issuing of new policies – will be a reduction of fraud. The large players like Mass Mutual have decades of data on which types of policyholders or claims are most likely to be problematic. In addition to preventing the issuance of policies to ‘likely’ fraud perpetrators, insurance companies can use machine learning to spot fraudulent claims as they come in. This will save countless hours of adjustor time and drive down wasted resources spent on false claims.

  2. Excellent article! I wonder how blockchain will, if in fact at all, fit into the ecosystem MassMutual has built. I think one of the biggest areas of opportunity is with the gig economy. With an estimated size of 68 million people [1] the gig economy is one of the largest opportunities in the insurance industry. In response to your first question, I believe the insurance industry will be one of the biggest resistors to entirely phasing out human advisors. Insurance products can be difficult to understand and the process requires a lot of specialization, which lends people to build personal relationships with a trusted advisor. It will take a lot of time and convincing for people to abandon their advisor.

    [1] https://money.cnn.com/2017/05/24/news/economy/gig-economy-intuit/index.html

  3. This summer I underwent the experience of applying for an individual life insurance policy, advised by my dad who is a broker in health insurance. I was shocked to see how antiquated the practice was. As you describe, it involved a nurse coming to my house, mailing of paperwork, mailing of checks, and an entirely unautomated system. I am thankful to hear that insurance companies are starting to narrow this technology gap! To your question of the role of human advisors in insurance – I believe that most of the administrative tasks will be replaced by machine learning technology. However, what cannot be replaced is the empathy required to enroll an individual in a life insurance policy – at the end of the day, what this policy is covering is the monetary value of your life, which I would venture to say most humans would rather have that be a conversation than a number provided by a computer. The deeply personal element of the industry will require the human interaction and empathy of humans, but can be improved dramatically in quality of experience by these newer companies. Thanks so much for posting!

  4. Very well research and thoughtful piece of article! With the rise of ML, insurance companies are adding new data points into their actuarial computations, which can help them underwrite affordable policies without the administrative overhead of highly paid human actuaries. We may see the rise of no-frills policies that can vastly improve lives of low-income groups that could not afford a traditional policy package.
    I am not sure if I would be comfortable with insurance companies crawling through my social media feeds though, and the use of big data to assess insurance premiums could raise privacy concerns. Will my insurance premium go up after one too many pictures of me gobbling up pizza? There is a limit to how much information people are willing to share with their insurance providers, and it will be interesting to see how this space evolve.

  5. Very well-researched and thought-provoking article! Every time I move apartments and am forced to go through the process of soliciting new property insurance quotes and calling insurance agents who ask endlessly redundant questions, I think about how ripe the industry is for disruption. It certainly seems like there are ample opportunities for machine learning to improve upon this process. However, when it comes to making decisions about offering coverage and determining premiums, I wonder if machine learning is better suited to some forms of insurance than others (i.e., property vs. life insurance). I’m reminded of our discussion in class about Amazon’s fiasco with its “accidentally sexist” hiring algorithm. Is there a risk that a machine learning-enabled platform tasked with making life insurance coverage determinations based on health and demographic information might arrive at conclusions that could be viewed as sexist, discriminatory, or otherwise undesirable?

Leave a comment