The case for AI and ML in insurance
The case for AI and ML in insurance
My previous post talked about the importance of data to artificial intelligence and machine learning. We see the most success when enterprises apply that data to three aspects of AI and ML. In manufacturing it enables real-time production improvements. Healthcare uses it for predictive analytics and modeling that leads to customized care. As promised, this post will focus on all three tracks in insurance: real-time, predictive analytics and modeling.
In manufacturing, AI and ML are more internally focused. You shouldn’t see it if they’re doing it correctly on the assembly line and getting the best products out there. Possibly you see it in end-to-end healthcare. Insurance puts it out in the open. You may have even seen commercials about one AI-based service: pay-for-what-you-use insurance (it’s done in life insurance as well as property and casualty).
Connected devices are just the beginning
Property and casualty insurers, for years now, have offered apps and devices that monitor your driving and send your data to the insurer for analysis. The insurers use the data to customize your rates based on your driving patterns.
Life insurers use data about lifestyle, age, and other data they collect, to similarly tailor a policy that’s unique to my needs, my goals, and my price range. The actuarial process, how they do underwriting and policy support, and everything else is unique to me.
The problem with life insurance in general, is that actuarial tables put us into certain buckets and risk categories automatically and if you deviate a couple of percent you wind up either up or down. What if I could connect my patient health record, my fitness tracker and my smartwatch, and send all their data to underwriting. The insurer would see that I exercise 3 to 5 times a week, so the fact that I may (or may not) be 15 pounds overweight would have a smaller impact on my risk categorization. So I get a different premium for my life insurance policy, or a different policy.
From there, as AI and ML soak into more processes, insurers can create far more value.
Microinsurance – Coverage is now possible on very small segments, which can be totally automated, from underwriting to settlement of the claim.
On-demand insurance – Coverage can be provided for a smaller, more punctual, more limited risk. Accelerated underwriting capacity is essential here.
Peer-to-peer insurance – Redistribution of a contract’s benefits to humanitarian causes, the shared purchase of a contract submitted to an insurer for the best price, or the pooling of the deductible.
Customized pricing – Perform the risk assessment individually through advanced analytics, machine learning and connected/wearable technology.
Changing the insurance value chain
I worked on an assessment recently for a classic, non-digital insurance company using neither AI nor ML, and generating about $1.5M in revenue per employee. It needed the capabilities to deliver data-driven insurance customized to the needs of its individual insureds. Without those capabilities it would lose market share to new digital competitors that were born with modular product and service structures, automated processes, underwriting customizations, and general lack of legacy baggage.
As an aside, this assessment raises another important point for your own AI and ML program. Make sure you have a framework to quantify the value of AI and ML in terms of market share, new services they can offer, and revenue and margin growth. Those are the big 4 results to track.
The insurer is infusing machine learning across six areas of its value chain:
- Device management
- Network management
- Device data capture
- Customer experience and management
- Real-time analysis
- Partner and ecosystem management
Those six areas are driving improvements to all of the insurer’s business services, i.e., product and service development, marketing, sales and distribution, policy management, underwriting, and claims and benefits management.
The insurer’s digital transformation is still underway, and it’s already realized the value of new services. It’s offering the microinsurance, on-demand, peer-to-peer and customized pricing I described earlier. At this rate, it’s on track to achieve the results forecasted in the assessment, which are typical for AI and ML in an insurance company:
- 15% market share growth as a result of digital channels
- 20% reduction in costs due to automation
- 5% reduction in costs due to risk and fraud management using AI and ML
- 87% increase in revenue per employee (my favorite metric)
That last one is an interesting metric because when we look at companies that use AI and ML correctly, the revenue per employee increases by between 50 and 70 percent at a minimum. They become more profitable. They generate more revenue. They have better services, and they can offer more diverse services.
What AI and ML are teaching us
As you consider the use cases for your organization, I hope my own learnings and experiences will help you along your digital journey.
If you haven’t already looked, here’s another link to the research we commissioned in partnership with Google Cloud: https://atos.net/en-na/lp/ai-readiness. The AI-Readiness web page contains a lot of information on AI and ML. A note on usage: If you register for the IDG Research/CIO Magazine webinar replay, you’ll receive a PDF copy of the MarketPulse survey report.