AI has made an impact for simple claims but data management limitations make it a longer term strategy for more complex claims.
Straight-though-processing (STP) is an oft-quoted benefit of adopting AI to validate claims, detect fraud and speed up settlement. Yet even the darling of the insurtech world, Lemonade, finds itself dragged into human intervention on anything more than the simplest of claims.
Homo Sapiens are a complicated animal. On the one hand immediate satisfaction is demanded but on the other they want to know that a fellow human being is at hand to empathise and sort things out.
And take fraud detection. 360Globalnet has processed over 2.5 million claims digitally for insurers and guess how many home claimants withdraw claims during the FNOL process?
See the answer at the bottom of the article. For now a digital FNOL process that enables self-service input.
- A brief text summary of the damage/loss/theft followed by
- a pushed link to add photos and then
- pushed link to add video yields surprising results.
Opportunistic claimants that over-egg the loss, or make a completely fraudulent claim see their claim gradually unravel and a high % voluntarily withdraw to avoid being found out (again see % end of article). Simple claims like: -
- Red wine stain on carpet
- Damaged TV (typically just before World Cups)
- Lost or stolen phone (typically just as new iPhone or Android model released)
It's hard for AI to spot these so let the claimant dig their own hole freeing up the claims pipeline and reducing claims costs.
For serious fraudsters? Well they just change the MO to beat AI every time. It needs augmented intelligence- the combination of technology and human experience and intuition. 360Globalnet uncovered 27 crime rings orchestrating auto fraud for one UK customer. Amazing the concentration of bent legal firms, 2nd hand auto dealers, breakers yards in close proximity! Accessing, joining and analysing claims and other data (structured and unstructured) reveals legal forms with suspicious claims patterns and joins up the links with claimants and dubious supply chains.
It's not that AI is not important it's just that you might rest too many hopes in fast and compelling solutions.
Unless you can already already capture raw (unstructured) data in a simple way to enable AI tools to utilise in a more productive way then you are not ready.
- Self-serve FNOL
- Through customer and supply chain orchestration
- Via Real-time status updates for customers and staff
- To settlement
Then add the data management capabilities which helps you be AI ready. Make sure the platform's API architecture lets you add the AI solution of choice as an integrated part of the total claims solution.
That is why McKinsey puts a long lead time on adoption across more than simple and specific use cases.
Still, should be a major strategic part of vision and strategy so time to revise the business vision and strategy to exploit the benefits of AI.
AND the ANSWER? 35% consistently quarter after quarter.
Claims. Carriers achieve straight-through-processing rates of more than 90 percent and dramatically reduce claim-processing times to hours or minutes, from days. Internet of Things sensors and an array of data-capture technologies, such as drones, largely replace traditional, manual methods of first notice of loss. Customer interaction with insurance-claim organizations focuses on avoiding potential loss. While no one can predict exactly what insurance might look like in 2030, carriers can take several steps now to prepare for change.