In the season finale of Insurance Unplugged, Lisa Wardlaw wrapped up this six-part series with a focus on some of the trends that are impacting insurers as we move into the new year.
Lisa talks with Helen Yu, Founder and CEO of Tigon Advisory Corp., a fractional CXO-as-a-Service growth accelerator that multiplies growth opportunities for companies at any size. Helen is also an author and the host of the CXO Spice podcast, focusing on AI, digital transformation and more.
In this episode, Helen and Lisa cover a lot of ground, from how to know where to start with AI in insurance to how to prepare for a successful implementation. Here are five takeaways from the episode.
The Importance of Knowing Where to Start
Start with your domain knowledge.
When it comes to insurance, there are four key domains where AI can play a key role: underwriting, claims processing, risk assessment, and lastly (and most importantly) customer engagement. When you think about insurance and AI transformation through this lens, then you can really think about where you are, where you want to be, and where you need AI to help optimize these pillars.
You have to understand your data really, really well. What type of data do you have for each domain?
Is it policy data, claims data, customer records, industry-specific data sets? Next, identify where you collect the data, why you collect the data. Do you have any exposure or risk? When you’re training AI model you need to know where any potential exposure lies.
And then identify a business problem. Where do you have the biggest business problem? Is it in your internal underwriting process or the claim? Does the process take too long, or do you have major risk exposure from a cyber perspective or from customer engagement. Do you have a low customer experience rating? Are you losing profit from premiums? There is a reason for that. Perhaps you can automate that process, have AI do more fraud detection, initially, and then pass it to the real agent to prevent fraud.
Once you have done this groundwork, then you know where to start.
Setting Up for Success
When you know where you’ll be working, the most critical part is to benchmark where you’re at.
Think about how you are going to measure your success. For example, you may want to reduce claim processing from a week to one day. If you don’t have a benchmark, you may never know how much to invest and what it really takes to generate the results you want.
Then, don’t forget to build in the resources, the skillsets you need to make it happen because this is also part of your investment.
Finally, think about what will happen if you don’t do anything. To me, this is the biggest risk for companies: by not doing anything you could disappear in a few years.
Upskilling: Domain-Specific Expertise
Vertical expertise will absolutely become more highly valued in the era of AI because you need people with the technical background who can create relevant prompts and also understand insurance. This domain knowledge and the ability to connect the dots between the many areas within this domain is absolutely critical. I’m talking about connections between tech and insurance, and from those areas to risk management and customer engagement and to compliance and regulation. By connecting these dots, you’re better able to collaborate with partners and customers as well.
Humans Are Essential for AI
Some people have the misconception that AI will replace them. The reality is, AI doesn’t have the knowledge. AI doesn’t know the regulatory requirements for insurance until you build that into your algorithm, and we really need humans in the loop to remove the bias. AI is biased to begin with, so humans have to be in the loop to take that bias out of the data. The higher the quality of your data, the better modeling you can build. To build that model without bias you have to have humans who represent the majority of all humans, not just a single race or a single gender.
The 3Ps for AI in Insurance
When it comes to AI in insurance, there are three Ps that you have to think about.
You must have a purpose, in terms of the purpose that you want to provide to the community and in terms of how you plan to leverage the technology.
Next, is the problem you’re trying to solve. You don’t just want AI for the sake of claiming that you have AI.
Finally, you have to have patience. AI is a tool. You have to put in the right data and have the right representation of the data, allow that to really run and test it. You have to train the AI so it knows what you want to achieve. You have to interact with the AI to allow it to work more efficiently and in a way that serves the greatest good for people and society. This takes time and patience.
Listen to the episode with Helen Yu on the Insurance Unplugged podcast.