Tab navigation


Our panel of industry experts dove into three main areas:

  1. How today's innovation trends in society are viewed through the lens of machine intelligence
  2. Machine intelligence's impact on (re)insurance from a value chain/market perspective
  3. The organizational implications from the rising adoption of machine intelligence

For the purposes of this discussion, 'machine intelligence' was used as an umbrella term for the following:

  • Artificial intelligence: Computer systems designed to undertake tasks that usually require human intelligence.
  • Artificial general intelligence: Computer systems that can perform any intellectual task also performed by a human (in some cases becoming self-aware).
  • Machine learning: Computer systems that can observe and learn tasks rather than being explicitly programmed.
  • Deep learning: A type of machine learning where a system can learn from data without prior assumptions based on models or underlying frameworks.
  • Cognitive computing: Computer systems that simulate the processes in the human brain (which can then become one basis for machine learning).
  • Augmented intelligence: Computer systems that assist humans in a set of intellectual tasks or tasks that support intellectual tasks.
  • Expert systems: Computer systems that use databases of knowledge to provide advice.
  • Robotic process automation: Tools and systems that can replicate or "roboticize" repetitive back-end processes.

The current explosion of data — it's a fantastic thing, but not without issues:

Things are moving faster than expected and it is imperative that individuals gain control over their data. Who owns and controls data is important. It can empower an individual. 

One challenge with AI is that when you transform it into business execution, it becomes challenging. To eliminate cultural barriers, there's a need to bring together new data workers and the business/actuarial community.

It's also worth remembering we have been here before with respect to AI. In the early '90s, there was a flurry of interest in AI based on progress in applying neural networks. For a variety of reasons, the potential was not realized, and so an "AI winter" dawned, which only thawed in the last 10 years. In the past two to three years, we have seen another round of hype (in the Gartner Hype cycle, this is called the "peak of inflated expectations"). We have now slipped into the "trough of disillusionment" for machine-intelligence related technologies.

Will we enter a second winter, or will we make it to the "slope of enlightenment" that leads to productivity?

Questions to ask ourselves: What are the sign posts, and how should we plan and prepare for different trajectories of machine-intelligence related technology adoption?

You want to use all the data available to you. Great, but first consider:

There are barriers getting to the first step, and algorithms are a commodity. The problem is getting data in the right format from scanned documents, 10% of data is structured and 90% is unstructured. Ultimately, it is important to make sure the business case for AI is solid enough.

In some ways, data is the new oil. There's the need to extract it (data collection); there's a need to refine it (data curation); and there's a need to distribute it (data visualization).

Good data beats algorithms, and data preparation is more important than algorithms. But it can be a challenge to extract good data from documents, and millions of data points are required to understand risk. Of course, the more you collect will result in better data-driven outcomes.

About those insurance-value chain opportunities:

Use data to come up with better pricing and better interaction with clients. What is key is how the information is received by the individual and how it changes the way they can behave, e.g., 'this is what it will do to benefit you.' It is important to gather information and share it back in a partnership model to change behavior. You need to be comfortable in describing how you can change behaviors for mutual benefit.

One challenge today is the data 'black box,' which is problematic for financial-industry regulators. The industry needs to be better aligned so that regulators get up to speed on how the industry views data. This is challenging because there is little sharing of info among industry players, we are still in a proprietary mindset.

On implementation:

You need to find the right business case, and then prepare change management and transformation. You need subject matter experts and adequate effort to evaluate all options. As with anything, adoption is fundamental to make it succeed.