The Aware Machine in insurance

Create a vendor selection project & run comparison reports
Click to express your interest in this report
Indication of coverage against your requirements
A subscription is required to activate this feature. Contact us for more info.
Celent have reviewed this profile and believe it to be accurate.
31 July 2015
Michael Fitzgerald
The topics of artificial intelligence, machine learning, deep learning, and cognitive computing have made their way into the popular business press. An insurance professional trying to stay informed of emerging technology may struggle to see the application of these technologies to their industry. A Celent new report, Machine Intelligence in Insurance: Designing the Aware Machine, provides an explanation of this space and its opportunities in insurance. It describes a model named "The Aware Machine", identifies the characteristics of high-value problems best suited for such platforms, and suggests specific use cases to serve as proof-of-concept experiments. The use cases include:
  • Analysis of legal circulars for impact: Continuously identify which regulatory changes will have a material impact. Involves teaching a system insurance law and providing it with a continuous feed of changing regulations.
  • Medical case management: Optimize treatment plans to increase recovery, return to work rates
  • Identification of underwriting leakage: Analyze insurance contracts at the clause level and compare them with each other across lines of business to enforce consistency of intent. Continuously monitor new contracts to ensure that appropriate wording is used.
We invite readers of this blog to submit their own candidates in the comments section and check back for updates. Let's crowdsource suggestions and get some proof of concept experiments underway!


  • With P&C carriers diversifying underwriting operating functions are often a challenge. Centralized or field seems to drive org discussions. The applicability of learning machines, AI and Robotic process automation should help alleviate the desperate submissions into a single stream of data along with third party information from various sources such as Fitbit or fuel band IoT aware devices. Underwriting can be re-engineered again to focus on the exceptions and have the promise fulfilled of issuing complex risks quickly. I believe this will also apply to term life with restrictive limits and data access.

  • Hi Mike,

    An interesting example might be classifying normal and abnormal behaviour in operations teams. Abnormal behaviour might signify the new rising stars, fraudulent activity or the rise of a new trend the company needs to respond to. The trick is, getting the leaders in the organisation to wade through all that activity and look at the interesting bits - a use case that might apply to a learning system.


  • Hi Mike,

    To kick this off, here's another potential use case...

    Road traffic accidents: Analyse the images, locations, traffic data, climate data, vehicle data and claimant details to identify the causal relationships and apportion blame, with the objective of learning / adapting over time and settling the majority of claims with minimal intervention by claims assessors, CSRs, controllers and supply chain partners.

    Parts of this use case have been done already, however not as a single integrated learning system.



  • Mike ... I think that distribution would be a candidate. Determine the penetration rate within segements and efficiency of channels.

    Most insurers have some form of this in place, but the aware machine would have a broader spectrum and would be able to continuously refine the criteria in near real-time, allowing for more effective scenario planning and performance monitoring.



Insight details

Insight Format
Geographic Focus
Asia-Pacific, EMEA, LATAM, North America