The Human Touch

Cognitive computing

There was barely a pulse on an LED between the victory of IBM Watson on the American game show Jeopardy! in 2011 and the hype beginning. The cognitive computer system had just defeated two former winners of the show in a ‘man vs. machine’ playoff for a $1 million prize and, according to the media, it was now poised to reshape the world.

In 2013, IBM announced that the first commercial application of Watson would be for managing decisions concerning lung cancer. Since then, Watson is said to be forecasting weather, writing recipes, tackling development in Africa, and working on a smart assistant for General Motors. Last year, it even wrote a song about heartbreak - “It’s Not Easy” - with Grammy-winner Alex Da Kid.

“Watson is probably not quite as ubiquitous as it seems,” says Roy Loeschner, Global Head of Shared Services at Allianz Global Corporate & Specialty (AGCS). “But there is no doubt that cognitive computing will change the world.”

Science fiction is filled with tales of intelligent machines functioning autonomously from humanity, but reality is more mundane. Many experts believe artificial intelligence is impossible; instead, we are now at a stage referred to as “cognitive intelligence”: humanity leveraging computing power to enhance its own abilities. According to Loeschner, this technology is already changing the insurance industry.

In 1809, Napoleon intentionally made an illegal move three times during a chess game. His opponent, a machine known as the Mechanical Turk, then swept all the pieces from the board. Amused, Napoleon sat down and played a real game – and lost in 19 moves. Considered a wonder, the Mechanical Turk proved to be an illusion controlled by a human chess master. The desire to create artificial intelligence clearly has a long history. It was not until 1997, however, that the world saw how artificial intelligence could outsmart a human. IBM Deep Blue, a chess-playing machine, defeated reigning world champion Garry Kasparov in six games.

The Jeopardy! victory of Watson over two human champions 14 years later was another breakthrough. Until then, it was said that machines were good at doing what humans were bad at. Computers were faster at calculating and processing, but a disappointment at understanding natural language or recognizing unique objects in an image. Watson’s victory showed that machines could now do what humans are good at. Watson, which can process information equivalent to a million books per second, solved cryptically worded questions, revealing an ability to understand natural language. The milestones have continued to fall ever since.

This year, Google’s AlphaGo triumphed 3-0 over the world champion of Go, China’s ancient abstract strategy game. “During those games, AlphaGo made a completely new, and thus, creative move which had never been made by a human player before and which the system had not been trained on,” says Daniela Schneider, the Allianz Group expert for text-related artificial intelligence.

She asserts that the AlphaGo victory is significant because of the complexity of the game. Played on a 19x19 grid, Go has more possible configurations than atoms in the universe. “We are seeing a future where computers are beginning to solve problems in entirely new ways. Until now, creativity has been considered an innately human domain,” says Schneider.

Google’s AlphaGo triumphed 3-0 over the world champion of Go
Played on a 19x19 grid, Go has more possible configurations than atoms in the universe. On the way to a 3-0 victory over a human champion, AlphaGo made moves never seen before.

Self-learning algorithms and data mining, as well as pattern recognition and natural language processing, are combined in cognitive computing to mimic the way the human brain works and learns. Where cognitive computing distinguishes itself from humanity is in its brute processing force.

A medical researcher can, for example, read half a dozen papers a month.A cognitive computer reads about a million in 30 seconds. In addition to speed, it can also read test results and doctors’ notes to discover patterns in data and recommend the most promising treatments to use. “Such capabilities,” says Loeschner, “assist in fields drowning in documents and data.”

Current projects at Allianz that utilize artificial intelligence include fraud detection, human speech recognition and robots with natural language abilities to process claims. Based on a recent test case in India, the use of robots has already cut normal processing times of submissions from 25 minutes to 3 minutes or less. This represents a productivity gain of 85 percent, and all made 100 percent error-free.

Daniela Schneider: "Until now, creativity has been considered an innately human domain”

At the moment, the technology is mainly used for process automation. “This is the low hanging fruit of the technology,” Loeschner admits. “They bring immediate efficiency results, enabling us to handle claims faster, provide new business quotes, issue policies and improve the customer experience.” Several Allianz operating entities are, however, assessing further application areas for cognitive computing. Loeschner points to the work of Schneider and her colleagues as another area where innovation is occurring.

Schneider, who has degrees in literature, linguistics and computer science, is working on text analytics – projects that retrieve relevant pieces of information from unstructured content to enable greater insights and thereby support better decision-making processes. Allianz is ramping up its capabilities in Natural Language Processing to be able to read and understand documents in multiple languages much faster than any human.

Another project aims to identify global supply chain risks. Take the example of the 2011 Thailand floods, where unprecedented flooding caused disruption to many of the biggest companies in the world. Many of the companies affected were not even aware that their supply chains extended into Thailand, so were taken by surprise when components stopped appearing. Intelligent solutions could have helped identify these risks well in advance.

With the help of Natural Language Processing, publicly available information is now screened daily for information about supply chain relations between companies. “The information is visually depicted, so underwriters can assess the risks,” Schneider explains. “It also enables them to understand and to advise the client better.” Praised by both industry partners and clients, this Allianz application based on Natural Language Processing is now used globally by underwriters.

Roy Loeschner: "there is no doubt that cognitive computing will change the world."

The biggest impediment to cognitive computing is data – where do you get it from? Cognitive computing is highly reliant on having masses of data to analyze. But for many insurers, this is difficult to deliver. At AGCS, for instance, cognitive computing is not immediately practical because the company does not have the volume of data required. “As an industrial insurer, we have thousands of customers, not millions, so we don’t have the data load where it makes sense to run cognitive computing projects ourselves,” Loeschner says.

Where cognitive computing is being used is within Allianz-wide initiatives, or in combination with external providers. One example is AGCS’s recent partnership with Praedicat, an insurtech analytics company, to predict with increased accuracy the upcoming key catastrophe liability risks. Such advancements in analytics will redefine the role of underwriting and claims management, says Loeschner.

Schneider agrees. She expects underwriting to be transformed by cognitive computing. Powerful cognitive computing tools will support underwriters in assessing risks with greater accuracy.

“I believe cognitive computing will also revolutionize customer service by optimizing communication and reducing response times.” For most people, interacting with an insurance company is a negative experience in their life, particularly when they are making claims relating to losses, deaths or disasters.

“This will never be an enjoyable experience,” says Schneider, “but technology is assisting us to make it a far smoother process that puts the customer and their needs first.”

As with all content published on this site, these statements are subject to our Forward Looking Statement disclaimer:


Gregor Wills
Allianz SE
Phone: +49 89 3800 61313

Send email

Apr 18, 2024

Allianz completes transaction to sell its 51% stake in Allianz Saudi Fransi to Abu Dhabi National Insurance Company (ADNIC)

read more

Apr 18, 2024

AllianzGI receives approval to commence wholly foreign-owned public fund management business in Mainland China

read more

Apr 17, 2024

How are Allianz employees playing a role in the Olympics?

read more