You’re in Good [AI] Hands: Machine Learning Implementation at Allstate
Insurance companies are increasingly relying on machine learning to create cost efficiencies. Allstate is seeing success in its machine learning implementation to date, but what are the implications for its employee base and long term competitive advantage?
It’s no secret that the insurance industry has relied on mountains of data since its inception. Risk underwriting inherently relies on complex calculations that benefit from tens of thousands of data points to appropriately price policies and build algorithms. In fact, Allstate has been using elements of AI to routinely price personal lines of insurance for 20 years.[1] However, now Allstate and its competitors are leveraging new technologies to extend machine learning (“ML”) beyond actuarial science to the areas of claims adjustment and customer service, large cost centers within traditional insurance businesses. In the case of Allstate, these technologies alone will not be long term competitive advantages, but will be a necessity to compete with its peers and other disruptive startups in the space, such as Lemonade, an AI focused insurance company which raised $120 million in Series C funding in December 2017, led by Softbank.[2] The competitive advantage, however, will be derived from Allstate’s future ability to integrate its ML technologies, its proprietary data, and its human workforce.
In September 2017, Allstate implemented IPsoft’s AI platform, Amelia, to serve as a ‘digital colleague’ to human customer service representatives. In contrast to completely automated chatbots, Amelia works with a human rep to generate possible solutions using natural language processing and data analytics. When Amelia doesn’t know the answer to an inquiry, she will observe the human rep’s response and learn from that interaction.[3] According to Allstate, in the first six months since deploying Amelia, average call duration decreased from 4.6 to 4.2 minutes, and inquiries solved in a first call increased from 67% to 75%.[4]
Regardless of how powerful the standalone ML technology may be, it will always require some degree of human input, and organizations are more likely to be successful when they view ML implementation as upgrading human roles, rather than cutting them, even if workforce reductions may be the ultimate outcome.[5] To date, it appears that Allstate is taking the appropriate steps to make sure the human and ML integration is successful. For one, its CEO, Tom Wilson, appears to have the right philosophy, recently saying in an interview “[From the $500 million of savings from tax reform] we put $40 million into increased training because we’re worried about the growth of artificial intelligence and what it’s going to do to service jobs…We have to figure out how to train people to do the new job, not the job that the computer can do”. [6] Further, 99% of Allstate agents working with Amelia had expressed complete satisfaction as of March 2018.[7]
While Allstate is putting the right training in place to ensure near-term success of Amelia’s implementation, it needs to be open with its employees about their long-term role at the company and provide them with a sustainable employment path. As of June 2018, Allstate had already cut approximately 1,000 claims personnel, largely driven by ML capabilities. [8] This number will grow as ML technologies become more efficient. What responsibility does Allstate owe to its employee base that is becoming increasingly overstaffed in light of ML advances?
Further, as it becomes less expensive and less sophisticated to service claims with a high degree of customer service when aided by ML, how will Allstate maintain a competitive advantage? Whereas the experienced sales force and customer service of Allstate may have been a differentiator in prior business cycles, Allstate is coming to realize that its ability to monetize more data than smaller competitors may, in fact, be the go-forward competitive advantage. [9] While it appears that Allstate has a cohesive ML strategy to date, the tension between ML, data, and its human workforce will only grow in the years to come.
(Word count: 792)
[1] Tom Wilson, interview by Erik Schatzker, Bloomberg Markets: European Close, Bloomberg, June 28, 2018.
[2] Jonathan Shieber, “SoftBank leads a $120 million round for insurance startup Lemonade,” TechCrunch, December 19, 2017. https://techcrunch.com/2017/12/19/softbank-leads-a-120-million-round-for-insurance-startup-lemonade/, accessed November 2018.
[3] “Allstate Elevates Customer Service Through Artificial Intelligence,” press release, May 23, 2018, PR Newswire, https://www.prnewswire.com/news-releases/allstate-elevates-customer-service-through-artificial-intelligence-300653613.html, accessed November 2018
[4] Sara Castellanos, “Allstate’s ‘Digital Colleague’ Amelia Answers Questions For Call Center Reps,” CIO Journal (blog), Wall Street Journal, March 30, 2018, https://blogs.wsj.com/cio/2018/03/30/allstates-digital-colleague-amelia-answers-questions-for-call-center-reps/, accessed November 2018.
[5] Jeanne Rose, “The Fundamental Flaw in AI Implementation”, MIT Sloan Management Review vol. 59, no. 2 (2018): 10-11, ABI/INFORM via ProQuest, accessed November 2018.
[6] Tom Wilson, interview by Erik Schatzker, Bloomberg Markets: European Close, Bloomberg, June 28, 2018.
[7] Juan Martinez, “Amelia and Allstate in the Wall Street Journal”, IPsoft press release, March 30, 2018, https://www.ipsoft.com/2018/03/30/allstate-and-amelia-a-perfect-match/, accessed November 2018.
[8] Sarah DeWitt et al., “Allstate: Mgmt Meeting Takeaways: Upbeat on the Opportunity for More Profitable Growth”, J.P.Morgan, June 15, 2018, https://amr.thomsonone.com/, accessed November 2018.
[9] Ibid.
As Allstate continues to grow its ML business, I believe it will be crucial that it continues to offer training opportunities and retain its existing employee base. As a consumer that places a premium on customer-provider relationships with such products, I feel that maintaining the human element is crucial and serves as a point of differentiation. Additionally, as other companies within the industry continue to cut staff, this maintained focus on keeping that human customer experience should provide for an increased sustainable competitive advantage.
I find the application of Amelia to be incredibly interesting. The fact that they are able to use machine learning in a part of their business that is so integral to the overall company is extremely advantageous. I do think that eventually Allstate will have a moral dilemma on their hands, but from a business point of view it is a good issue to have. As long as Amelia is able to continue to leave customers more satisfied than a customer service rep, and Allstate is then able to use this as a way to lower overall cost to the customer, I think the Allstate will remain a large player in the insurance game for years to come.