Q&A: Nationwide CTO Jim Fowler talks generative AI

Fowler:  AI is not a new term in our world. For the last decade, Nationwide has deployed predictive models of artificial intelligence to help our teams make life insurance underwriting decisions and process claims. What is different for us now is the emergence of generative AI and large language models.  

We are witnessing the coming of age of large language models that can scan vast amounts of data and generate predictions. With ChatGPT4 technology, it can create content based on an extremely large knowledge base. The other piece is natural language processing, which is the ability to ask and answer questions in natural language using the large language model to generate content.  

The thoughtful integration of generative AI will enable companies to build stronger relationships with customers through a more personal, reassuring, and effortless experience. That is where our focus is.  

We have spent the past five years digitally enabling our products and services with modern integration platforms and intelligently automating processes. This work is pivotal and allows us to plug into the ecosystem of our intermediaries. Today, we are at the start of a new phase that I call augmented intelligence. This technology is going to empower a bionic enterprise that bring humans aided by machines to run the company and serve customers, agents and financial professionals in a vastly different way for the next 100 years. 

Fowler: Nationwide’s approach to generative AI is driven by our AI Steering committee.  This group includes leaders from across our enterprise that guide our engagement and balance innovation with responsible development. The goal is to help prepare our employees to use AI in smart ways and find and evaluate use cases to support our business.  

Additionally, we have what we’re calling a Blue/Red Team risk approach. Our “Blue Team” explores how we can use this technology to benefit Nationwide, the ways we can harness its power to help our associates be most productive, provide drafts of work faster, and ways to better serve our members. We also have a “Red Team” that anticipates the potential pitfalls of generative AI. They consider issues such as cybercriminals misusing the technology, unethical use for malicious purposes, and how to prevent bias from affecting decisions or work driven by the technology.  

Companies that approach this correctly do not only look for ways to benefit from generative AI, but they also think about all the possible risks that come along with it and address them proactively. 

Fowler: The combination of generative AI and a human mind brings speed, efficiency and unique problem-solving benefits to businesses. By allowing employees to harness the power of generative AI, we give them a virtual assistant to automate simple tasks. Not everybody needs to be a technology expert to use this in their job every day, and we continue to learn as we go.  

I think the jobs of today will change and evolve because of this technology. There will be a different set of skills needed going forward. The underwriter of the future will not just need to know how to underwrite – the skills will shift, and they may need to train a bot or build a model to help underwrite a policy. They may even need to manage those bots as virtual employees.  

Another benefit that we see from a generative AI perspective is that it gives us new ways to think about solving problems. It helps us find an alternate route that we had not thought about – and it challenges us to rethink the way that we might approach a problem or issue.  

In all cases, we have set guardrails, and we believe there must be a human in the loop. Our bionic world is humans plus machines – we want to make sure that in any decision that is made specifically with generative AI, there is a human involved in vetting the output of the model. People in the loop to provide empathy, judgment and reason needed to complete a project, the machine doesn’t have that capability.  

Fowler: Generative AI is changing the way that everybody does their work, and this will keep evolving.  

We recently launched Bing Chat Enterprise, which gives all our employees access to OpenAI’s GPT-4 (the technology behind ChatGPT) but for internal use in a safe and secure way. We are seeing many early examples of how this technology helps our employees be more productive. We believe generative AI can help with most jobs by producing the first draft of the work product.   

For example, we are testing generative AI for first generation of code development and legacy code conversion. Instead of having a developer set up their code, they use generative AI to generate a first draft. Then the developer can improve it, add more complex components and ensure it meets our standards, such as security. This speeds up the code development process.  

Our pet team also used generative AI to help create our Pet HealthZone, a site that provides health information for pet owners. The first draft of all the content was developed using the latest generative artificial intelligence, which was trained with our content. Our expert veterinarians on the team took the first draft from this technology, refined it, and then transformed it into valuable content for our customers. 

Consumers are using generative AI models to research investments. They use it to compare different companies and investments from different sources. They use it to evaluate the pros and cons of investments.  We think it has potential to be a great tool for advisors, too. Imagine an advisor who has a specific customer walk in their office. For example, a single adult, age 50, who wants to retire in the next five years and wants to know what to consider when buying an annuity? Generative AI could search all the different sources for that information and bring back the most relevant ones to help the advisor prepare their proposal for that specific customer. We see that level of personalization in financial advising as essential.  

With any of these scenarios we believe that a human must remain in the loop. There is still too much risk and incorrect information in these models to rely on them without keeping a human to provide the actual advice. We also believe that judgement, empathy, and critical thinking skills are vital for advising and these skills are uniquely human.