This means that customers of IAG brands, which include NRMA Insurance, SGIC and SGIO, involved in a motor total loss accident will get a claims outcome faster and back on the road in a new car sooner.
The technology, which combines artificial intelligence with business process automation, has helped IAG achieve up to a two and a half week reduction in claims times for customers when their car has been written off in an accident, by removing the need for a vehicle to be towed to a repairer prior to being assessed as a total loss.
IAG Director of Analytics Hannah Sakai said predictive total loss, which was developed in-house, was created to help reduce the emotional impact of a car accident by providing customers with more clarity and certainty sooner in the claims experience.
'A car accident can be a traumatic and challenging time for our customers, so we turned to artificial intelligence to help improve this experience.
'Our predictive total loss solution leverages machine learning to detect a potential total loss with more than 90% accuracy, using information provided by the customer when they make a claim on the phone with a consultant or online.
'The customer is notified of the potential total loss outcome via text message the following day, providing transparency upfront on the process and providing answers to commonly asked questions. We've seen a significant uplift in customer advocacy as measured through total loss customer experience surveys.
'The predictive total loss model is one of many AI applications being developed by data scientists in IAG's AI Centre of Excellence. Using an internal team allows us to leverage our unique business knowledge to tailor the experience to our customers', Ms Sakai said.
How predictive total loss works
Predictive Total Loss automates business processes to deliver proactive and transparent customer communications that keep customers informed at each stage of the motor total loss experience. It removes manual processing steps to settle customers' claims sooner.
In addition to the positive impact to customer advocacy, Predictive Total Loss has:
Put customers at the centre of the design process by reimagining the total loss experience to resolve specific customer problems and pain points when a customer has had a car accident.
Reduced claim cycle time and claim costs, a sustainable uplift in claim processing efficiencies and productivity.
Automated aspects of the total loss claim processing, removing the need for our claims teams to perform tens of thousands of manual processes each month. This frees up time for our claims teams to focus more on helping customers and has improved overall efficiencies of the claims teams.
Application of AI ethics framework
Prior to deployment, Predictive Total Loss was evaluated using IAG's established AI ethics framework and the Australian Government's voluntary AI ethics principles to identify potential issues or risks prior to go-live, including:
Human, social and environmental wellbeing: making sure the objective of the project was to benefit IAG's customers, with no other conflicting objectives, and clearly documenting this to assist with ongoing monitoring.
Reliability and safety: experimentation to verify that customers had a positive experience and setting conservative thresholds for modelling to help reduce the likelihood of wrongly predicted total losses.
Fairness: Careful consideration of the potential benefits and harms of the system, including the distribution of benefits and harms across the population.
IAG has joined a number of other businesses to test the AI ethics principles announced by the Minister for Industry, Science and Technology, the Hon Karen Andrews MP in November 2019.
The IAG AI Centre of Excellence plans to refine the model using customer photos of the vehicle damage and extending the methodology to predict motor claim liability to help automatically validate claims at the time of customer lodgement.
IAG - Insurance Australia Group Limited published this content on 30 November 2020 and is solely responsible for the information contained therein. Distributed by Public, unedited and unaltered, on 30 November 2020 04:14:05 UTC