The use of cognitive computing within international programmes

Technology is starting to play an important role in global insurance programmes, through the automation of knowledge work. There are two main aspects to this: robotics process automation and cognitive computing. While robotics mimics the actions that a human takes to do something, such as processing a document, it is not a thinking process. However, cognitive computing goes further by trying to mimic how a human brain operates in trying to understand meaning, in the context of what is being said.

Cognitive computing involves advanced AI algorithms, which allow the computer to read text (in any shape or form, whether handwritten, typed, email, text) but more importantly, to understand both the structure and logic of the text, and draw conclusions from it.

Cognitive computing and global programmes
International insurance is inherently complex. There are many different aspects to consider when issuing policies to cover an organisation’s global risks, such as different jurisdictions, market practices and standards, varying legal interpretations of insurance contract clauses plus the regulatory differences. It can be a challenge for an underwriter to take all of these different things into account and ensure that everybody understands exactly what it is that is being covered in the various countries.

The policy has to reflect the intent of the agreement between the customer and the insurer. There is an inherent risk in any programme that the global policy wording may not correspond with local policy wording. The only way to ensure full alignment is for the underwriter to read and check every local policy against what has been agreed globally. This is time consuming and becomes even more difficult when the local policies are in a different language.

So, there is an issue of time, and of capability, and this is where cognitive computing can play a role. When an underwriter tries to understand whether a local policy is fully aligned with the underwriting intent, they have to understand the policy, to know where to find things and how to interpret what they see, in both the master and local policy. So, for example, cover may be discussed in one section of the contract, with exclusions in another section, and exceptions elsewhere in the same contract. Cognitive computing can mirror that exactly by reading both documents, analysing them, holding them together and giving immediate feedback to the underwriter as to whether they are aligned or not.

There is also an educational benefit for underwriters, as they are now starting to see and understand things about countries that they were not necessarily aware of: issues that are highlighted by the technology. In this way, cognitive computing clearly supports the objective of contract certainty, ensuring the policies do exactly what was intended, and helping to speed up the whole process.

The current project
The process of creating a cognitive computing engine for global programmes is not a simple or quick one. It involves subject matter experts, those that would normally perform the policy reviews, explaining exactly what it involves, what parts of the contract they look at and how they find out if there is an alignment or not. All of this then has to be fed into the engine, and the outcomes have to be analysed and any corrections have to be made.

For us at Zurich, we started with two lines of business. Firstly, liability, as it is complicated and non-standardised and so presented a major challenge. If our solution could handle liability, then it could probably handle anything else. At the other end of the scale, we also started with financial lines, as for this line the master and local policies follow the same structure, so should in theory be easier to align. And we focused on policies in English, as it covers the largest part of our global and local policy languages. At the moment, liability is in production and is being actively used by our underwriters, and financial lines is in the final stage of testing.

For other lines, it is a question of time and money, since it involves a human teaching the computer the business rules and is very labour intensive. It is not an off-the-shelf technology. A complex problem requires a complex solution to solve it effectively, and global programmes are highly complex. But once the solution has proven its value, it can then be extended, to more lines of business, more countries and more languages. Ultimately, it should allow, for example, a master policy in Chinese to be compared to a local policy in English.

As to the future, the sky is the limit. The application of cognitive computing also encompasses underwriting, creating an automated method of issuing a quote instantly, and claims, by supporting claims handlers with the assessment of certain types of claims.

In the end, cognitive computing is the ultimate preventative support, and means that customers and underwriters can be more confident about their risk management. This is not about replacing people, but allowing underwriters to focus on risk analysis and customer value-adding activities, and less on remedial issues. Although it is still early days, we are confident that customers, brokers and insurers can all benefit from the arrival of cognitive computing in the global programme space.

Contributed by Michael Blattner, head of IP Academy, commercial insurance, Zurich Insurance Company, and Dirk Gawronska, senior project manager, international programmes, commercial insurance, Zurich Insurance Company.