Liberty Mutual CIO tackles digital transformation complexity

Facing complexity as part of digital transformation

What are a few of your biggest technical challenges right now?

James McGlennon: The first one is dealing with technology debt. Many of our solutions are 20 or 30 years old, or even older. We are on a journey to the cloud, and a big part of that is rebuilding new applications in a cloud-native fashion. In many cases, we are looking for ways to run the old workloads on the new platforms. That's a difficult challenge.

The second area we spend a lot of time on these days is what I call data wrangling -- how we prepare data to be used in our machine learning models and our AI models. We have lots of data going back over 20 or 30 years -- sometimes even longer -- and it's not all in the same shape.

The third challenge that we have technologically is making sure that everything can be API-enabled. As we think about how we can participate in global ecosystems with new types of partnerships and partners, it's very important that we can plug into their world and they can plug into ours, and it becomes a seamless kind of interaction model.

Including ethical AI  

In what ways have you started to deploy AI and how has it changed your business?

McGlennon: We began an AI journey maybe four or five years ago, really doubling down on our investments in machine learning and exploring how we could take advantage of that. We also struck up some partnerships with organizations like MIT, where we're doing some research focusing on questions like: Is it possible to build synthetic data to be able to test our hypotheses and models? Are we able to validate why a model gave a particular recommendation? In this new age of AI and machine learning, we'll need to be able to explain how the models came to their conclusions. And regulators and folks will be asking for that as we move forward.

We have lots of data going back over 20 or 30 years -- sometimes even longer -- and it's not all in the same shape.

We have hundreds of machine learning models in various stages of operation across Liberty Mutual. There's hardly a part of our business where we're not deploying some new machine learning models.

For example, we have AI in our operations for our technology infrastructure that uses machine learning models that can predict when we're going to have a problem. That way we can solve that issue with the platform or with the solution before it causes an outage or gives our customers a poor experience.

We've also deployed AI and machine learning in computer vision technology. If one of our customers has a car accident, they can use their phone to take some pictures of the damage of the car and upload those with us. We can feed those pictures into a machine learning model that we've trained on millions of photographs of damaged cars. It can help us detect the extent of the damage and make an initial repair assessment. In some cases that can tell us immediately if the car is a total loss, so we can resolve the claim as quickly as possible.

Another one that comes to mind is that we're using machine learning engine technology in our claims processing engines. If customers call us, we can intercept them with a bot and maybe give them the answer much more quickly than if they had to wait for a human being.

We're trying to figure out how we can build a set of technologies that will allow us to, if necessary, retrain the models. Sometimes if you deploy a model in production, the data that underpins it can change or drift over time. We need to be able to continually track and retrain the model if necessary. That means we're investing a lot in new technologies to automatically deploy machine learning models through API connections that allow us to do it much more quickly. So speed is key, being able to retrain the models across the entire value chain.

Addressing employees' need for flexibility

What changes did the pandemic bring to Liberty Mutual, and which ones do you think will exist in the foreseeable future?

McGlennon: We're focused on how can we make sure that [employees] stay connected. That was a big challenge for us, given that we were all virtual. We needed to introduce new kinds of formalized approaches to checking in with people to make sure that they were feeling part of the team and that they weren't isolated.

In today's world, we will need to provide more flexibility to our employees than ever before. We lead with keeping our employees safe and we continue to keep that at the center of what we do.

We have begun a journey to hybrid work. It'll probably take a little bit longer to get everybody comfortable with coming into the office a bit more often, but our main focus is making sure that people have the flexibility they need. We've been very successful doing the things that we need to do during the pandemic. We won't be bringing people back in just to check off the box of 'I was in the office.' It'll really be focused on the activities that people are doing and the need to get together for specific activities or specific tasks that are better in person.