In this AI + Leadership conversation, Vlad Barbalat, President of Global Risk & Capital Solutions and Chief Investment Officer at Liberty Mutual Group, and Monica Caldas, Global Chief Information Officer of Liberty Mutual Insurance, explore how AI demands a fundamentally different approach than past technology implementations. Led by David Reimer and Adam Bryant of The ExCo Group, the discussion reveals three key themes: treating AI as a relationship rather than a tool, accelerating adoption through agency and experimentation, and shifting strategic value toward judgment and portfolio discipline as routine work becomes automated.
Reimer: Companies have been rolling out technology implementations as part of their investment and capital strategies for many decades. What’s different now, as you’re having conversations about AI, technology, and the future?
Caldas: I’ve been doing digital transformations my entire career. What is different today is that most prior transformations were treated as individual projects, not as business reinvention. When you orient the body of work as a project, there’s a start, there’s an end, there is a relatively isolated group of people involved, and a technology shift happens.
Sometimes there was an operational shift that took into account the upstream and downstream process changes that were required to fully integrate the technology. What makes this moment different is that we’re approaching it as an ecosystem evolution. There is still a technological shift, there’s still an operational shift, but there is a cultural transformation that has to happen, as well.
Barbalat: I think people became accustomed to technology causing a bit of functional change—a new software package is going to allow me to do A, B, and C a little bit differently. And by the way, the tech department is going to come and install it on my computer, or I’ll download it from the cloud, learn how to use it, and once I master it, this will be my slightly modified workflow.
That is exactly what is not happening with this particular technology, because it is not a vertical technology. It is a horizontal technology. It requires you to treat it as an experiment, treat it as a relationship, and treat it as a version of an assistant or employee.
Unless you invest in that relationship, you’re not going to get much back. You need the mindset that this is not just another thing to learn, but rather it is a relationship to explore on an ongoing basis, with the assumption that all kinds of things will change. There will be all kinds of ingrained habits that will evolve. That is very, very different than anything we’ve done to date.
Bryant: The notion of it being a relationship is fascinating. It infuses the ecosystem, and the ecosystem itself is in flux. This is not a lever we’re switching from off to on. From an investment perspective, how is the investment in AI infrastructure similar, different, or unique compared to other large-scale investments of the past?
Barbalat: Given the speed and scale with which capital is pouring into all forms of this digital backbone build-out, some poor capital-allocation decisions inevitably will be made. That can be because a technology doesn’t deliver the ROI or doesn’t deliver the productivity you expected, or that it simply takes longer to be fully adopted than people had assumed.
The other classic component of such a moment is that there are companies that really have no choice but to participate. Even though the group outcome may be irrational, it is very rational for the individual companies. So if you’re one of the hyper-scalers, you don’t really have an option of sitting this one out.
You may be worried that there’s going to be overspend, but the risk of not participating is, frankly, greater. And your shareholders, stakeholders, and board will hold you accountable if you don’t participate. So you get this dynamic where so many investors are pouring capital into the sector.
As an investor, I expect that there will be some poor capital outcomes for some firms, but I’m also confident that the industry is building an asset that will be productive overall. Depending on how you invest, where you invest, and when you invest, people will have a dispersion of outcomes. But if you’re attentive, you can build a responsible portfolio amid all the hype.
Caldas: We can draw on history to understand this moment. Another example of infrastructure buildouts can be found in a pair of famous pictures that are easy to find online. One shows Fifth Avenue on Easter Sunday in 1900. The street is filled almost entirely with horse-drawn carriages. The same street just 13 years later is filled with automobiles.
Think about the infrastructure that needed to be built in those intervening years, including the roads, the lighting, and the governance mechanisms. We’re in a similarly transformative time, yet it will happen much faster than 13 years. So we’ve got to accelerate all the adjacencies in order to prosper with AI. We’re all rewiring how we work and learn and manage.
Reimer: How do you drive adoption rates of AI through the organization? What have you learned about the best ways to get people to jump into the cold swimming pool?
Caldas: I’ll start by saying that you don’t learn how to swim by reading about water. You have to jump in, and then you can discover if you have an innate ability or whether you need lessons. Part of my mission is to make sure nobody gets left behind. So we have to figure out the right mechanisms to meet people where they are, based on their different skill levels and learning styles.
We talk a lot about “Everyday AI,” which includes sharing modules with employees that show what AI can and cannot do. We also have metrics to monitor the adoption rate of our employees.
Barbalat: This is a team sport, so everything we’ve done within the investment organization here at Liberty Mutual has been in partnership with Monica. We have very high adoption rates, and we’ve also partnered with external AI labs on a continuous basis. So as newer features of the models show up, we’ve been quick to use them.
We’ve also worked hard to lead by example, including asking people which tools they used to produce a body of work. This technology brings tremendous knowledge and capability into everyone’s hands, but it’s the agency of the individuals—to harness it, to explore, and interact with it—that’s critical. We need people to have more agency, rather than just telling them to double-click on a particular feature in a software program.
Bryant: Some people are making the argument that AI will lead to growing democratization of functional expertise. So which leadership skills will become more important over time?
Barbalat: Some skills will remain important, including the ability to ask the right questions, which requires critical thinking. Unless you ask AI the right questions, you’re not going to get the knowledge deployed in the right way.
If you look at articles from when the calculator was introduced, many people were worried that math skills would decay. That didn’t happen, of course, because in order to use a calculator, you have to know what problem you are solving.
Caldas: Think about engineering, which is much bigger than writing code. You’ve got to architect the system and consider all these macro complexities around governance, security, movement of data, etc. So the engineering field is so much bigger than just writing code.
That said, we can use AI to write code, which will raise the baseline productivity everywhere on that particular part of the software development life cycle. That will allow people to spend more time using their judgment.
It shifts the strategic value in each function. Routine work becomes more automated, but engineering is not going away, and software engineers can shift to more complex work. I also think the ability to iterate and to experiment is an increasingly important leadership skill if you want to go fast.
The other skill I’ll mention is portfolio discipline—the ability to know what to prune, what to turn off, and what to double-down on as a potential winner. Because everything looks like it could work in the beginning. But you’ve got to have that discipline to know what’s not working, and when to shut it off and move on.
Reimer: How is AI and technology changing your job today, and how are you evolving the way you approach your role for the future?
Caldas: I am starting with what we’re trying to do as a company in the future—what our business aspirations are—and then working backward from that future vision to know what we need to have in place now in terms of how my team and I operate. How do we make that future a reality, rather than just shifting and tweaking what we do today?
It requires a different mentality. You’re not just extrapolating from today. I am using different mental models. And it requires even more alignment around capital allocation and value creation. Because there are so many places we can invest. What are the top priorities? What’s the ambition? And how do we work backward to know what to put in place to unlock that capability that we’ve never had before.
Barbalat: One thing that has changed for me is that I approach everything that comes my way through the lens of, did we deploy AI in some way to get to this conclusion? That has been a continuous question that I ask myself and I ask of my team every day. You’re constantly learning. You’re trying to synthesize information. And I can use AI as a kind of infinite tutor. That’s incredibly powerful, because it introduces agency into the equation.
One of the ways I try to digest information is I create a board of advisors or kitchen cabinet, except it is all AI-generated. These are personas, and I’ll pick a couple of great investors or great CEOs and ask them how they would approach a particular issue or deal.
The crazy thing is that if you do that long enough, you have to remind yourself sometimes that I’m not actually talking to Jamie Dimon. You get the point—it’s a way to socialize information and learn in a much more efficient way, which ultimately makes you better equipped to do your work.
Bryant: It takes a certain mindset and background to be comfortable in roles like yours, where there’s a lot of ambiguity, and you are figuring things out in real time. Where does that come from for each of you?
Caldas: Vlad and I have a common thread in that we are both immigrants who came to the country at a young age. My family took a huge risk to come to a place where we didn’t speak the language. You’re on a mission to figure things out quickly. You can’t be afraid to take risks, and failure was just part of life. You skin your knees here and there, and you have to keep moving. You learn from everything all the time, from every angle.
I look for knowledge in every angle today, and I’m comfortable saying I don’t know. It’s just how I grew up. And I am attracted to operate in cultures like this one that have an entrepreneurial spirit and are meritocracies.
Barbalat: I remember coming to the United States when I was nine years old, and I was a paper boy a week later, trying to figure out a way to contribute to help the family survive. That’s the approach you take, and you do that with gratitude. You don’t do it with a mindset of a victim. You go to work with excitement. You grab it. You’re grateful for the opportunity.
And then things just build on themselves. I go back to that word “agency.” You realize you can do things and be successful. You have this mindset of achievement, and you don’t spend a moment of failure on sulking or victimhood. You just continue to build. That’s what immigrants are forced to do when they come to the US. That said, there are other ways to be wired like that.
Caldas: When I think of all the amazing peers and leaders I’ve had a chance to work with, there are these common themes of being able to solve problems and also bring people along with you. They don’t have a self-centered motivation; it’s a win-together motivation.
You don’t have to be an immigrant to do this, but you do have to be wired a certain way to always want to work together to solve problems and not be afraid to put issues on the table and work through them, shoulder-to-shoulder.
Recent
