AI and Insurance: Are Claims Jobs in Danger?

Mirroring the actions of at least one large Western insurer, a Japanese life insurer announced it would introduce cognitive computing in its claims department in late December.

Executive Summary

In the first of two articles about artificial intelligence and insurance, Carrier Management reports on carrier efforts to use cognitive technology to replace or to work alongside their existing claims processing staff.

In Part 2, we aggregate research from various sources about the impact of automation on underwriters and agents, as well as on professions outside the industry—businesses they insure.

Unlike other insurers, however, Fukoku Mutual Life Insurance indicated that its new system would displace a portion of its workforce, pushing tweeters and bloggers into high gear during the New Year’s holiday break with links to a Dec. 30, 2016 article titled “Insurance firm to replace human workers with AI system.”

Details of the article reveal the number of displaced workers at Fukoku will be around 34—roughly 30 percent of the staff of the payment assessment-related department.

A Dec. 26, 2016 statement about the move from the life insurer is difficult to interpret, even with the translation provided by Google here. But the English translation of a related Dec. 30 article that followed in the Japanese National Daily Mainichi explains that the introduction of Fukoku’s artificial intelligence system will make the process of reading medical certificates more efficient.

The workers being made redundant by cognitive technology powered by IBM Watson will be drawn “primarily from a pool of 47 workers on about five-year contracts” rather than permanent staff, the article says. “Final payout decisions will continue to be dealt with by a dedicated staff.”

What Is Cognitive Computing?

Celent’s Craig Beattie distinguishes AI concepts like deep learning and cognitive computing in his article, “Deepening Automation and AI in Insurance.”

Noting that the life insurer is initially spending 200 million yen ($1.7 million) to install the IBM Watson Explorer system and expects to pay 15 million yen ($129,000) annually to maintain the system, expected savings from staff cuts are estimated at about 140 million yen ($1.2 million) per year, the Mainichi report says. “The company is planning to let a number of the contracts run out their term and will not renew them or seek replacements.”

Technology publications and blogs were quick to predict dire implications in the days following the Fukoku news.

“The human race thought it had a bit longer, but the end is in sight. Or at least it is in Japan, where IBM’s Watson has started taking our jobs in the insurance sector,” says the introduction to an article posted on telecoms.com.

An investment advisor, Mike “Mish” Shedlock of SitkaPacific Capital Management, started his analysis on his blog Mishtalk this way: “Manufacturing jobs have already been decimated by robots. White collar workers are next in line.”

2016 WEF Report: The Future of Jobs

In January 2016, the World Economic Forum announced the publication of a report, “The Future of Jobs,” with conclusions based on a 2015 survey of chief human resource officers, chief strategy officers and other senior talent and strategy executives from 350 of the world’s largest companies in 15 countries.

The report found:

As many as 7.1 million jobs could be lost through redundancy, automation or disintermediation over the next five years, with the greatest losses in white-collar office and administrative roles.
The loss may be partially offset by the creation of 2.1 million new jobs in more specialized “job families,” such as “computer and mathematical” or “architecture and engineering.”
Healthcare is expected to experience the greatest negative impact in terms of jobs in the next five years, followed jointly by energy and financial services and investors.

Even BBC joined the chorus on Friday last week, referring in its report to a study by the World Economic Forum released in January 2016—”The Future of Jobs.” The WEF study predicted “that the rise of robots and AI will result in a net loss of 5.1 million jobs over the next five years in 15 leading countries,” the BBC article says.

The 34 jobs on the line at Fukoku are being replaced by a “diagnostic document assessment automatic coding system [that] automatically encodes, classifies,…categorizes diseases, disasters, surgery, etc.,” says Google’s AI interpretation of the statement from the life insurer. While much is lost in translation, key functions of the IBM system will include: recognizing the name of an injury or disease described in the medical certificate or similar document submitted with an insurance benefit request; automatically extracting dates (hospital admission, surgery); and “learning the history of past payment assessment to inherit the experience and expertise of assessor workers.”

“[A]fter introduction, we will continuously reflect the final assessment result of the person in charge,” the statement says.

Identifying benefits, the statement says “we will improve the accuracy and efficiency of our operations” and that customer service will improve along with payment accuracy.

Efficiency at Zurich: A Trend Beyond Life Insurance in Japan?

Further west on the map, in Switzerland, the group chief operating officer of Zurich Insurance Group was clearer about how the property/casualty and life insurance company is using cognitive technology in a similar fashion.

“This is not about people doing a sloppy job. But when you read a 100-page report and you come to page 100, whether you still understand the linkages between page 1 and page 100, as a human you make mistakes.”

Kristof Terryn, Zurich Insurance

During an Investor Day presentation in November last year, Zurich Group COO Kristof Terryn gave a specific example of a project already underway in the U.K., in which cognitive technology is used to review medical reports for evaluating personal injury claims—reducing review time of 10-100 page documents from 58 minutes all the way down to five seconds.

This one project is expected to free up 39,000 hours of claim handler capacity per year, saving $5 million annually, Terryn said.

One big difference between the Fukoku system and the Zurich project: Zurich gave no specific indication that the group will cut staff as it introduces cognitive systems in claims and also automates risk assessment and underwriting functions. During his remarks to investment analysts, Terryn suggested only that the nature of jobs is changing as he described targeted process improvements and other automation initiatives aimed at increasing efficiency and technical excellence. (Separately, reports of a media conference held in connection with the Investor Day event suggest that while Zurich has embarked on a new plan to cut costs by $1.5 billion from 2015 through 2019, Zurich Group Chief Executive Officer Mario Greco has indicated that prior plans to cut or relocate 8,000 jobs as part of the cost-cutting activities have been scrapped.)

Describing a “focused set of initiatives to improve customer experience and…effectiveness in [the] claims” function, Terryn said, “We have various efforts underway to further push claims process automation through robotics and cognitive tools.

“Robotic desktop and process automation and cognitive computing will bring benefits in productivity. It will also allow us to reallocate work hours to much more value-added tasks.

“For more complex tasks, we’re focusing on cognitive technologies which enable natural language processing and hypothesis-driven decision-making to drive better claim outcomes,” he added.

An analyst who pressed Terryn for more discussion of exactly how the cognitive technology of the U.K. project will trim $5 million from expenses—or $120 per hour for the 39,000 hours of claims handler time now being done by computer—suggested that the savings must be loss savings rather than expense savings, coming from reduced claims leakage rather than from lower staffing expenses. “It’s clear to me that claims managers aren’t paid $120 per hour …So [are you] saying basically that the claims management staff are processing eight or nine cases per day and making quite a few mistakes?” the analyst asked, interpreting Terryn’s math to suggest that the performance of the claims team had been subpar.

Terryn responded that cognitive computing clearly “leads to efficiency” or lower expenses but that “the bigger impact longer term” is in fact the result of more accurate claims settlement costs. He did not agree with the entire premise of the analyst’s leading question, however. “This is not about people doing a sloppy job. But when you read a 100-page report and you come to page 100, whether you still understand the linkages between page 1 and page 100, as a human you make mistakes,” Terryn responded.

Part 2: “Are Underwriters, Agents Endangered by AI?” will publish on Jan. 10

“Computers have a better memory. When we train computers to see linkages between certain patterns in a claims file,…this is what we’ve seen. We’re now in the fourth phase of what we’re doing in the U.K. around cognitive computing, and what we’ve seen is the quality of assessment coming out of an automated process tends to be better than what we see from a claims handler.

“So, going forward, the upside we get from claims leakage—from claims settlement costs—may actually be better than efficiency piece” or claims expense savings, he agreed.

According to the Mainichi report on Fukoku, other Japanese life insurers (Dai-ichi Life Insurance Co. and Japan Post Insurance Co.) are using or gearing up to use cognitive computing, specifically Watson AI, alongside of claims staff rather than replacing humans.

Related articles: