Upskilling actuaries for the Age of AI

What does it really take to upskill actuaries for AI and modern data work? 

In this episode, Steve speaks with Cal Skwerski, an actuary who has moved from traditional reserving into analytics and technology-led transformation, about how Cal approached building data literacy at scale. The conversation explores why early training efforts fell short, how a shift toward a shared baseline of skills changed adoption and why community, peer learning, and hands-on “capstone” projects proved critical. 

A central theme is the importance of creating a common language around data, AI and governance enabling actuaries to collaborate more effectively across disciplines and manage new risks with confidence. Rather than focusing on tools alone, the episode offers practical insight into how culture and learning by doing can turn upskilling into a catalyst for real transformation.

Mixed & Edited by Next Day Podcast

info@nextdaypodcast.com

Transcript

Steven Abel 

Hi, this is Stephen Abel. I'm a global technology leader at Oliver Wyman Actuarial. And I'm here with one of our amazing leaders and practitioners, Cal, to talk about the wonderful world of technology and upskilling practitioners in AI and technology. Cal, do you want to tell us a little bit more about yourself?

Cal Skwerski 

Hi, I'm Cal. I'm an actuary on Oliver Wyman's data and technology team, and I help actuaries and other professionals working in the corporate and insurance spaces upskill with modern practices and data and analytical tooling. I've taken on the role of overseeing Oliver Wyman's global actuarial data literacy initiative, and over the past year, my focus has been on driving the modernization of the tooling and techniques our actuaries use and practice.

Steven Abel 

So tell us a little bit more about our journey at Oliver Wyman Actuarial. How did we start? Like how did data literacy become a thing at Oliver Wyman Actuarial and then maybe why?

Cal Skwerski 

Well, we started by identifying a need for some sort of organized and universal training. And that came partially from the success and high value of the actuaries we saw within our organization that already had high data literacy, but that was a small percent of our people. Soto motivate this a bit more, these were the individuals that first were familiar and proficient with a variety of different tools and techniques, scripting languages, data visualization tools,low-code tools, version control procedures, and cloud analytics platforms, this familiarity and experience allowed them to choose and use what was best for their jobs. 

Second, these highly data-literate actuaries were already capable of evaluating and reworking complex data processes, often simplifying procedures embedded in Excel workbooks, either requiring some degree of unnecessary manual effort to reproduce or processes that were computationally inefficient one way or another.

And third, these were the people that always were bringing forward new tools to try new ideas. They had read an article online about a new technique or new AI release and wanted to test that in practice and experiment in some type of safe sandbox environment. So we decided that really instead of this being the exception or the unicorn actuary, what we really wanted is to teach all our actuaries these skills and equip them with the resources they needed to thrive in this environment.

Once we had this vision for what our actuarial practice as a whole could become, we were able to get buy-in from leadership to start creating the curriculum and we already have plenty of momentum to roll out the upscaling to our practice.

Steven Abel 

So walk us through the journey. So, okay, there was this recognition that we have the introduction of AI, we have some of these new technologies who want to be doing more interesting work. How did we get started?

Cal Skwerski 

Well, the first thing that we did was we needed to determine what the gaps and opportunities in the data skills within the firm were and how we could develop data literacy and training that would address those gaps and lead to more opportunities. 

We had a hypothesis that there were certain pockets of expertise within our firm, such as predictive modeling expertise, scripting expertise, data visualization and actuaries and data scientists within our firm that were very capable of building dashboards or building products for our clients. What we needed to reveal was what percent of our practitioners that possessed those skills and how could we disseminate those skills more broadly across our practice.

So to do this, we developed polling in order to reveal the skills that we did have. And the polling revealed that this was mostly accurate, that there were certain pockets of expertise, maybe 10 to 20 % of practitioners actually possessed certain skills in scripting or in data visualization techniques or in usage of AI models more recently.

We also realized through this process that there were skills that we did not need to universally train on, that our actuaries were already proficient in Excel and our scripting. So this really shifted and allowed us to determine what we needed to focus on through this program.

Steven Abel 

So what did version one look like when we rolled out the training?

Cal Skwerski 

Version one, we actually trialed data literacy in 2024 with a small cohort of maybe six to 10 actuaries that was single skill focused. We rolled out data literacy with this program to focus on teaching Python to our actuaries and using Python in their day-to-day work. 

We had this thought that we would try to teach it like a university course almost where we had one to two experts acting like professors and our curriculum would go from basically this is the first time you've ever opened a scripting language, a notebook and here's what Python is and here's what a line of code is and try and extend that all the way throughout the course of eight weeks to now you're a proficient Python expert, data science expert, you can build models, can build APIs, automate processes, everything like that. And we failed remarkably and learned a lot from that actually.

Steven Abel 

I remember that we actually had a university professor that was an actuary that was working for us that designed that curriculum. Okay, so sometimes failure can lead to successes. What happened next?

Cal Skwerski 

Well, we learned a lot from that trial. And what we did was, first of all, we changed the curriculum to focus more on a broader set of skills that we wanted to establish as a baseline. So we were less concerned with drilling all the way into depth and developing a couple more experts in a certain field. 

And we were more concerned with establishing a universal baseline skill set for our 150 actuarial practitioners that would develop this universal data language and allow our actuaries to communicate effectively, work on similar things, share IC, and add that overall layer of baseline efficiency that we could then extend off of and develop more advanced, develop advanced training, develop more sophisticated intellectual capital for the practitioners that did have the latitude and the capability to explore things like rag models, for instance.

Steven Abel

Were there other than just the curriculum itself, were there other things that we did to make the second attempt at data literacy more successful for us?

Cal Skwerski 

Yeah, there were. So across the firm, we made the training mandatory, which was the first big spark. To get people to do things, usually it helps when you make it a requirement and you tie outcomes to that requirement. The second thing we did was incentivize and increase the recognition of the people who had, the practitioners who had gone through the program and successfully completed the program.

A part of that too was developing a platform and allowing our practitioners to share what we built and refer to as the capstones of the project, which is at the end of the program. The practitioners will develop either a tool or modify a program or something that they use in their day-to-day work in order to, that reflects the skills that they learned through the program. We allowed the practitioners to share their capstone work with the data community that we had developed and lead advanced training sessions on some of the work that they had done.

Steven Abel 

Did the community help at all? And in a lot of training, you take the training, you learn a lot, and then you kind of forget it almost immediately. Was there any aspect to that?

Cal Skwerski 

I think the community helped a lot and, you know, even myself, I surrounded, or I was able to build up a community of practitioners that I found that had skills that I didn't or were able to practice the trainings that we had taught in a different way to develop, you know, a new type of program, a new dashboard, a new Alltricks workflow that they took and automated a process that was repeated 50 times across the practice. And now we're able to bring and share that within our community and others were able to benefit from that intellectual capital that we developed.

Steven Abel 

This sounds great, but it can't have gone perfectly. Can you tell us about some of the challenges that you've encountered or we encountered in this journey?

Cal Skwerski 

We were met with a lot of challenges in the journey. First and foremost there was an element of cultural resistance to the requirement of mandating and training or requiring a lot of time from people who are already really busy. Part of this also is lent to by the diversity and skill sets that the practitioners who would be taking this training possessed. There were limitations in how much time we could reasonably require from people. But as we began sort of rolling out the training and having people participate. 

We realized that the people who already possessed certain skills actually rose up and were really interested in helping the others too. So it was powerful to be able to leverage those internal champions and use their support to overcome this challenge of cultural resistance and instead of requiring them to do some of the same trainings that they already possessed proficiency in, we were able to actually use their skills and their will to help educate others on those skills.

Steven Abel 

So were there any sort of unexpected outcomes or surprises from this journey so far?

Cal Skwerski 

One of the most unexpected outcomes was our first trial with the program where we had really smart technologists leading isolated training. We found that that did not work effectively. And what did work was peer training and practitioners talking with other practitioners and through that community that we talked about. 

So not only having our experts and our facilitators leading training and supporting the, like let's say the assignment work or the creation of the capstone projects, but using someone who had already gone through the an earlier cohort of data literacy and gone through the same series of training and using them as a first point of contact to help another practitioner go through the training and complete their work.

Steven Abel 

Tell me more about these capstones and why they've been effective. What is a capstone and how is that relevant for the training and maybe the person doing the capstone?

Cal Skwerski 

Yeah, the capstone is my favorite part of the curriculum that we developed and I think where we see the biggest, the most benefit overall to both the students, the pupils and going through the training and also as the leadership group where we're seeing returns already. 

What we do for the capstone is we require at least a four hour investment for the practitioners to take the skills that they've learned. So it has to be something related to all tricks, Python, Power BI or Databricks. And we were asking them to use those skills and either create something new or modify something so that it enhances the efficiency of something in their day-to-day jobs. 

Basically just put their skills into practice and then write a small report on what that looks like and then share that with the rest of the group. So what we found was it was really enabling. I think it was the enablement and the requirement of you have to just break ground on something. It doesn't have to be a complete project. 

You don't have to boil the ocean but we're expecting you're using your new skills and you're able to produce something of value, almost like an MVP for something that can then either be invested further into or shared with others.

Steven Abel 

So learning by doing and learning by sharing is particularly important.

Cal Skwerski 

Precisely, yeah. The model that we found and then sort of evolved into through the program that worked really well is as a program and as leaders, what we're trying to do is enable the practitioners to allow them to develop. And if we can create that enablement, if we can provide the training that gives them the skills necessary to develop the second stage of this. 

Then we have a platform and are allowing and enabling our practitioners to develop new things and work on interesting work. And then we're providing them with the platform to share that work and disseminate those skills and that information and those products across the practice that enable, develop, share model is starting to compound really quickly. And we're really starting to see returns.

Steven Abel 

That's fantastic. Okay, how has this changed you and your role?

Cal Skwerski 

Personally, I found that these skills and the ability to communicate with our other practitioners through this universal data language and the skills that we've developed together has made my work a lot more interesting. I've been able to work on and sustain a wider variety of projects as well as the infusion of new technologies, new tools, the things that we're working at the forefront of in developing AI tools and building tools that are using, not only bringing a wider breadth of insights and available information to our clients, but also drilling into the depth of the data science world and the tools that we are able to develop to share a deeper level of insights with our clients. 

That variety of work and the depth of the work that we're able to go into now through data literacy has personally made my job much more interesting in the long term. And as a leader I've found that the skills that we've brought to our practice and through data literacy, are also realizing, people are also realizing that they can accomplish the same thing.

Steven Abel 

I'm curious about, I want to double click on this idea of a common language. So sure, it's important for a consultancy to have a common language to be able to do more work together. What are your thoughts around just actuaries and the actuarial profession having a common language around AI and technology? Is that, is that relevant to the story? What are your thoughts on that?

Cal Skwerski 

Well, it's critical and it's I'll extend that a little because I think it's not even just actuaries and as actuaries are sort of a cornerstone in the insurance industry as in terms of their roles and capabilities. I would extend that statement to. You know, the role of the actuary is changingand the scope is expanding on what actuaries can do and where they're inserted. 

So whether that's talking with and collaborating with risk managers or underwriters or other actuaries, that ability to communicate technically with those audiences and with IT and more technically proficient experts in the insurance field, actuaries are establishing themselves and evolving as the centerpiece of everything going on with AI and technology right now in the insurance industry.

Steven Abel 

That's interesting. Let's expand your thoughts on AI in general. It seems like everybody is getting exposed to AI across the board. Do you see this need for a common language, a lexicon for AI? And is there also a need for a common lexicon for other classes of technology? just curious your view on that.

Cal Skwerski 

I think there is a need for a common language for actuaries and for again practitioners in the industry, whether that's actuaries communicating with actuaries or actuaries communicating internally within an insurance carrier or inside of a corporate risk agency. And what that looks like could be different depending on again the company or the organization. 

But what we found is vital in progressing our transformative journey in data and AI is the, again going back to the enablement of the tooling and the languages and the practices so that as we're going through and our practitioners are learning how to use AI, they're learning the same, they're learning how to use it in the same way. If we're building, if we're building rag models or we're developing a series of prompts to perform the same tasks for a certain portion of our, of our certain procedure that there's consistency and governance within thatprocess is also vital. 

So the things that we're focused on training our actuaries and our practitioners when it comes to AI are how to establish consistency in prompting, governance, and the usage of our AI tools.

Steven Abel 

So that's interesting. You mentioned governance, and I often think about the actuaries role in helping manage risk. Do you think it's important to have a common language, a common understanding of some of these tools as we modify and extend our governance, particularly around some things like AI?

Cal Skwerski 

I think that the actuary and every professional has a deep responsibility to understand the tools that they're working with in general and data literacy can help with that through an educational process that helps people using AI understand what the risks of AI are, what proper use looks like and what misuse looks like is vital to a healthy ecosystem.

Steven Abel 

So Cal, what advice would you have for others? This has been a super interesting discussion about governance and AI and upskilling. But if someone's listening to this, what should they take out of that?

Cal Skwerski 

For others, I would recommend to start small, whether that's building a program for 150 actuaries, 500 actuaries, or just yourself. Start small, trial and find out what works for your organization. Build momentum, find a way to measure impact. And through that process, there will be positive fallout.

I think the biggest learnings that we gained were early on in understanding the approach that we needed to take to overcome common pitfalls that we were running into, to motivating people, to understanding what our program needed to look like in order to be effective. And I think that looks different for everyone.

Steven Abel 

If there are three things that we should take out of this conversation, what would they be?

Cal Skwerski 

The first would be that there's so much value in this program and the process of doing this and that investing in others is something that will pay off a hundredfold. The second is just to be flexible through this journey, through this process. There were a lot of things that we learned that didn't go right the first time or that we had to modify to meet people where they were at in order to be able to have this be something that was productive for them as well as our leadership group. 

Third would be just start the journey. Don't be afraid that it's too much or that you're going to have to change down the road. Ultimately, just taking the first steps is what was most important to us getting this underway.

Steven Abel 

Thank you. If someone wanted to continue the conversation with you, how could they reach you?

Cal Skwerski 

Your audience can reach me on LinkedIn or reach out to me directly at my email, cal.skwerski at oliverwyman.com.

Steven Abel 

Thank you, Cal. It's been a pleasure.

Cal Skwerski 

Thanks Steve.