Other parts of this series:
Should a cardiac surgeon be repairing an airplane’s overflow pressure valve?
What if an insurer could underwrite complicated accounts with just a handful of questions and no human intervention? Or underwrite life insurance through the analysis of an uploaded “selfie” instead of an extensive application? Thanks to Artificial Intelligence (AI), several companies now are using these and other customer-focused innovations.
The promise of AI has hit the insurance industry offering both enormous gains and worrisome consequences. Accenture research shows that 75 percent of insurance executives expect AI to transform or bring significant change to the industry over the next three years – and why not? What company wouldn’t want its brand attached to useful information triggered when consumers ask a question of Alexa, Amazon’s intelligent personal assistant? On the other hand, who would want to engage in a very public AI project only to find it subverted by users teaching your chat bot to spout unacceptable language.
Insurers are struggling to respond to these new technologies. Looking at some of the successful, highly publicized AI projects, they realize AI can bring important gains and that they need to be doing something with AI. Still, if they move too fast or in the wrong direction or fail to act at the proper time, consequences could be major.
So, what are the most common mistakes insurers are making? Here are five:
1. Adopting a project instead of an overall strategy
When the pressure is on to show rapid technological advances, insurers often move quickly to do a stand-alone project. This could be good, provided they have thought through where this will fit within the overall company goals. But if they rush to accomplish something – anything — without planning, they could wind up with minimal gains compared with high costs.
Larger scale can mean bigger benefits. Often technologies can be purchased and reused for different operations, spreading costs over a larger part of the enterprise. These economies require planning across many parts of the organization, and they require adhering to the enterprise strategy. But they can avoid creating orphan projects that will be expensive to maintain and eventually redone to fit future architecture.
Though IT should always be done strategically, the additional complexity of artificial intelligence automation has given this added importance. Ultimately, the entire process needs to fit together end-to-end, so task hand offs between machines and humans need to be smooth. This requires adhering to a broad strategy.
2. Trying to force-fit AI into everything
If your holiday gifts this year include a new and exotic spice, most cooks would check out recipes that used that ingredient in an authentic dish. If it were sufficiently exotic, though, the recipe would likely call for ingredients not on hand, and the cook would be left sourcing those ingredients or leaving them out or making substitutions—perhaps making it more of a burden than a gift.
It’s a similar situation when an insurer becomes infatuated with an interesting type of artificial intelligence development – or even with AI itself — and searches for a place to use it. As with the exotic spice, this approach will likely involve more than you bargained for and may never be a good fit. Instead, if you identify a significant problem or opportunity and then search out an appropriate AI solution, you are far more likely to deliver the results you anticipate.
As we test out more AI applications, it becomes clear that machines are good at some tasks while humans are better at others. Certain aspects of customer service, for example, can be handled very well by intelligent machines, but disgruntled customers usually should be handed over to a live customer service human whose natural emotional intelligence and empathy can diffuse the situation. It has also become clear that work that requires creativity is best done by humans.
3. Expecting your internal staff of generalist IT employees to drive AI.
If you were in charge of airplane maintenance, would you send in a cardiac surgeon skilled at replacing heart valves to fix a faulty outflow pressure valve? Of course not. Nor would the hospital administrator hire an aeronautical engineer to replace patients’ heart valves. All valves are not the same. Neither are all technology project. Preparing and executing successful AI transformations requires specialized skills and experience that most companies lack in their IT department.
At the same time, the demand for Artificial Intelligence engineers has exploded, with technology giants like Amazon hiring large numbers of AI talent from the relatively small pool of trained workers currently available. AI training programs have sprung up in colleges and universities and even businesses, while individuals with strong IT, math and physics skills are returning to school to acquire the skills they need to fill valuable AI positions. In some cases
Without the expertise to properly plan and execute an AI strategy, most companies will not achieve the AI effectiveness they seek.
Insurers are committed to AI. Accenture research shows that 75 percent say AI will transform or bring significant change to the industry over the next three years.
I’ll continue this topic next week, discussing the last two insurer mistakes when they move into artificial intelligence.
Learn more: Read, Accenture’s Technology Vision 2017 for Insurance
Accenture’s Technology Vision 2017 for Insurance
Read the 2017 Technology Vision for insurance--Technology for People--which identifies emerging technology trends across the insurance landscape.