Jodie Wallis, the global chief analytics officer at Manulife, which operates as John Hancock in the U.S., spoke with Digital Insurance about how the company is deploying Generative artificial intelligence. Wallis oversees the company's AI teams.
Responses have been lightly edited for clarity.
What has the evolution of AI been like at Manulife?
In 2016, we started focusing on machine learning solutions. Some early use cases we focused on were around intelligent lead generation. So, helping our sales teams and our distribution teams understand where the opportunities were and then helping them zero in on customers who needed our help and needed our services the most.
Over the years, we expanded into fraud detection and some insurance-specific, underwriting-specific activities. Then at the end of 2022, like most people, we were surprised with the incredible attention that came from the ChatGPT announcement and roll out. From there we added to our machine learning work a lot of Generative AI work, and we've been on that path for the last two years. I'd say the thing that we've realized most is that when you combine classical AI or machine learning with Generative AI, that's where you get the greatest impact from artificial intelligence.
Why has Manulife decided to make GenAI implementation a strategic priority?
I think everybody, including us, said there's some incremental value to be gained from using machine learning over and above traditional automation or statistical techniques. We worked hard at that, and I think delivered some really good results over the years. Gen AI was really a game changer. We looked at it as technology that had universal applicability to transform how work gets done. With machine learning, we were able to do some things much more effectively because we were relying on pattern detection. Now, we are able to take AI and put it into essentially everything we do. It's helping us serve our customers better. It's helping us streamline our operations. It's helping us accelerate our ambitions.
What gives Manulife an advantage in implementing GenAI?
I think the first thing is, we've been in the AI game since 2016, so we had a really good core group of data scientists and machine learning engineers that were able to understand generative AI very quickly and start to apply it. Second thing is, we made some significant investments in our cloud, data and analytics platforms, and we may not have realized the benefit that would give us in GenAI. We made those decisions and those investments because it helped us scale our classical AI work, our machine learning work. It also helped us create better insights from our data, but it really gave us a boost, because we have all of our enterprise data in the cloud, all of our data scientists work in the same AI platform, and the Generative AI tools are available to us in that very same platform. So, we feel like we were able to move from experimentation to deployment much faster than others that hadn't made the same investments in that technology platform and that data platform over the years.
We did that for many reasons. We did it for the strategic ability to to scale our AI capabilities. We also did it because we wanted a cost effective platform. What we didn't realize is that by moving to the cloud, we would get complete, easy access to kind of the world's best, large language models when they hit early in 2023 so it kind of doubled the benefits of the investments we had made.
I like to say, 'Oh, I knew this was coming, and this is why I suggested we go that way.' But the reality is, I don't think most of us knew exactly what was coming, and sometimes, the decisions you make, the strategic decisions you make really pay off. And for us, in this case, it did.
So, the way we're implementing GenAI, we've decided to look at deploying in a disciplined way and across three dimensions. First, we're looking at those use cases that have measurable value for the organization. Second, we're looking at those use cases that can be built once, once and deployed many times. So, we're very interested in, instead of specific solutions for very specific problems, what are those GenAI capabilities that we can launch in one part of the organization and then scale across many other businesses? And the third is, we're really trying to think about GenAI, not as just like, plug it into little points across a workflow, but how do we use it to reimagine the end-to-end workflow?
We're very interested in those use cases. Sales is a perfect example, and we did a demonstration of one of our earliest solutions, when we had investor day in Hong Kong last summer. We could provide slightly tweaked leads, or we could provide a slightly tweaked set of insights. But what we did is we looked at the end-to-end customer interaction workflow, and we said, 'How do we infuse GenAI across the whole workflow?' The feedback from our agents has been tremendous, because it's not just sticking something into an existing process, it's reimagining the process with the knowledge that we have these GenAI capabilities.
The other thing we've been doing is we've been very focused on saying we're going to get the best from AI if we're combining traditional AI and GenAI together. Now, from a business perspective, that's not meaningful, but the way we solve it, it's very meaningful, right? So we can best judge where we can get the benefits out of each type of AI most effectively and efficiently, and then bring them together and present that as integrated solutions.
The other thing we've done in our deployment is we're very focused on adoption, so I don't think this is a matter of building a bunch of GenAI solutions, and people will magically know how to use them and embrace them. I think a lot of people figured out how to use ChatGPT because they're interested and they like technology. They wouldn't have found themselves on that site if they weren't interested in technology or wanted to know more. But not everybody's interested in technology or wants to know more. However, we still want them to be able to participate in this economy. So we've done a lot around training and training programs, a lot around adoption. But where I think we've had the greatest success is this idea of customized prompting workshops. So rather than generically provide training to 38,000 people, we divide them up into groups, into cohorts, and we focus the workshops on what they do during the day, rather than on generic prompts or generic prompting. We've had lots of good feedback from that approach, and we'll continue to do that. We call them promptathons.
How do you upskill or reskill workers to use GenAI?
And I'm quite proud of this. When we launched our GenAI program, we decided that we were going to be completely transparent with everybody in the organization. So, we created, on our intranet, a GenAI hub. And on the GenAI hub, we put learning resources. Of course, we told people what we were doing, but we put out a call for ideas, and we maintained our GenAI idea inventory open to all employees. The ideas that they've generated and others have generated, how they move through the pipeline, pipeline of prioritization and development and deployment. And from that broad set, we have launched almost 30 solutions into production to date. So that's the first thing that we've done, and I mentioned the prompting workshops, we're doing them on all levels. We have specialized training for most senior leaders. We just did a session with our top 130 leaders in the organization on prompting for leadership. And then we have customized workshops for any groups that want it, but also we're trying to get kind of coverage across all of our teams. And we created a version, an internal version, a secure version of ChatGPT, which we call ChatMFC, and we made that available to all employees as well.
Nobody was born an expert in GenAI. None of us have 20 or 30 years of experience with this tool. So I feel like we're starting from the same spot, whether you are an AI person or not. So we went right out of the gate in 2023 and started providing upskilling and training for everybody in the organization. I think others that I've talked to have debated, you know, should we just start with this group or this group? And our approach was to say, 'No, we're going to start with everybody, and everybody deserves to have this training, to get upskilled and to be kind of ready for the next phase of work, whatever that looks like to me.' That is the key here to make sure that people have the opportunity to continue to grow in their careers and do more and more fulfilling things.
Is hiring a concern?
I read an article in the Wall Street Journal recently which said, banking is the new hot industry for data scientists. I'm hoping insurance will be next. Who would have thought banking, right? And what are the reasons? Well, the reasons are because of the investments that banks are putting in, and we feel like we're making those investments too. We invested in the platform. And one of the key benefits of the platform initially, not the accidental benefit we got later on, was it is great for recruiting. When we tell data scientists we have a completely cloud native platform with all the AI tools you need, and an integrated and integrated data environment that is something that they can't get everywhere. So we absolutely use that as a recruiting tool.
I think the way in which our senior leaders have leaned into AI, particularly in the last two years, has also been a great recruiting tool for us. I think data scientists don't want to go to an organization that says, 'Here you go. You can sit in this room and do things, but I don't want to talk to you because I'm just doing this and I don't really care about it. They want to hear the CEO is on his quarterly earnings calls talking about this.
Can you speak to the benefits of deploying GenAI?
I think customer experience is one area that is greatly helped by GenAI. But there's other areas too, right? There's definitely revenue benefits. There's definitely cost benefits. And to the extent that we can help all of our employees be more productive, it means we can do more as an organization, for our customers, for our shareholders, and for our employees.
How do you see GenAI fitting into the digital transformation for Manulife?
GenAI is definitely a key lever in digital transformation, not the only lever, but it's a key lever. And most of the initiatives we have under our digital transformation umbrella do have some aspect of AI in it. Our digital transformation is about things like improving our straight through processing rate, which we've gone from 68% to 88% over the last six years. It's also about improving our net promoter score, and we've made some major improvements.
We're also doing other things on the digital front, like improving our digital properties, making updates to our websites and our apps. And we're doing things like making our products and services easier to find and easier to use.
Any comments you would like to share about broader AI trends?
GenAI is here to stay, right? We're not going to focus on it for two years, then go back to what we were doing before. I think that it's a trend where a lot of the engineering work and data science work in organizations is really coming together. So up until GenAI, we kind of had engineers doing apps and data scientists doing models. And what we've learned now is, in GenAI, these things come together. So, I think we are going to see some changes to how organizations bring together their deployment capabilities, and how data science and engineering become much more integrated and linked. So that's one of them. I think the other thing is, we've been talking for a very long time in the media about democratization of AI or citizen data science, and I don't think that those promises really ever came to achieve the vision. I think now we have much more of an opportunity to start democratizing some of the AI development and we were just talking about allowing people who previously weren't data scientists or couldn't really do data science now participate in an economy that is super exciting and much more available, much more approachable.
I think the language that is used to discuss insurance, the story gets easier to tell with GenAI. The other thing about the way we sell insurance is we're selling insurance to help people live a longer, healthier life, and to promote longevity, and to explain how longevity actually works and what you need to do and how we can help. And that's not everyone's language, right? Insurance agents and brokers haven't been speaking that language for 150 years so GenAI is helping them to understand concepts and then explain them in a way that customers can understand.
