Saket Saurabh:
Hello everyone. Thanks for listening to another episode of Data Innovators and Builders. I’m your host, Saket Saurabh. Today I’m with Santiago Guillotti. Santiago is the Chief Data and Analytics Officer at Chubb. Santiago, thank you for chatting with me today. Let’s start with your background.
Santiago Guillotti:
Nice to meet you, Saket. Thank you for the invite. I think we have some great topics to discuss.
Saket Saurabh:
Tell us how you came to your current role.
Santiago Guillotti:
I currently lead the data and analytics organization for Chubb in Latin America. I have the pleasure of leading around 130 people across different roles: data engineering, architects, data scientists. I’m really proud of what they’ve accomplished over the past three years.
Before this, I worked at more tech-oriented companies, e-commerce and retail, and had the opportunity to build those teams from scratch. My background before that was in finance and insurance.
Saket Saurabh:
Awesome. Tell us a little about Chubb, the insurance business and the data side at a high level.
Santiago Guillotti:
Chubb is an insurance company, one of the biggest in the world.
Saket Saurabh:
One of the topics we were discussing before the recording was Gen AI. And you mentioned something called Gen BI. Tell us what you mean by that and what you’re trying to accomplish.
Santiago Guillotti:
There are tons of projects and initiatives around Gen AI, but one of the most interesting ones for large, mature companies is how to improve overall efficiency. There’s a significant opportunity in business intelligence-related tasks. A mature company like the one I work for now, or other insurance and banking companies, typically has around 35-40% of FTEs working on BI-related tasks. The opportunity is huge. With the new stacks and platforms available today, you can substantially increase efficiency in those areas.
Saket Saurabh:
You’re right, a lot of new technologies are coming in, a lot of innovation broadly. From what you’ve seen or tried yourself, what technologies are you finding most exciting, and how are they changing things for you?
Santiago Guillotti:
About new tools, this might sound less innovative, but the modern data stack itself is still quite new for large enterprises. These companies often still have their data pipelines on-prem. So when you talk about things like Databricks or cloud-based products from Azure or AWS, those are products these companies are actively adopting as they migrate processes to the cloud. And a key challenge is understanding how that cloud infrastructure and microservices connect with on-prem systems during migration.
Saket Saurabh:
The data stack has changed a lot and I’ve been seeing that firsthand. One thing I was curious about: you mentioned building data teams from scratch. Tell us a little about that process.
Santiago Guillotti:
The approach depends on the type of company. Building from scratch in a large, mature company is very different from doing it in a startup or a new business unit.
In a mature company, the data and analytics function is often the transformation agent for an old stack. The team needs to be resilient. They need to understand they’ll be working with legacy technology while trying to migrate it forward. In a startup or new business unit, you build everything from scratch using the latest tools available.
What you look for in team members in each scenario is quite different. And this might sound obvious, but it’s really important to understand what you’re looking for when you’re hiring. I see a lot of hiring processes that don’t clearly define that, and as a result, they struggle to find good candidates. It’s not just matching a job description to a CV. You need to understand the specific skills the team requires, beyond whether someone can build a data pipeline or train a model.
Saket Saurabh:
That’s a great point, it’s a bit like the Moneyball approach. You need different skill sets and strong capabilities in specific areas, and you need to think about how those people come together as a team. One question I had: you’ve built teams in Latin America, and many of us in the US are increasingly looking at building global teams. How is hiring, training, or building up a team in Latin America different from the US?
Santiago Guillotti:
It’s absolutely about culture. You need to understand the culture of the different regions. Working with people from Mexico is not the same as working with people from Thailand. In different regions you see very different approaches to problem-solving and to the interview process itself.
Because they approach interviews and work differently, you need to adapt your hiring process accordingly. I’ve hired people from Asia before, and it’s extremely different from Latin America. The key is: first, stay clear on what you’re looking for. Second, build a hiring process that actually surfaces that. And critically, don’t waste the candidate’s time. I’ve seen strong candidates drop out of processes for exactly that reason. The culture in Latin America is very different from Asia or EMEA, and you need to adapt.
Saket Saurabh:
Can you give an example of that cultural difference?
Santiago Guillotti:
In the US, people are more direct. They go straight to the point, and interviews need to reflect that. In Mexico and Latin America broadly, conversations don’t work the same way. If you come in as an interviewer being very blunt and transactional, you can be misread as cold or too tough, and that’s not necessarily what candidates expect or respond well to.
Saket Saurabh:
So you can come across as intimidating and miss the cultural nuances entirely. On the team-building side, what are the key differences between building a 10-person team versus a 100-person team?
Santiago Guillotti:
When I was younger, I assumed that a bigger team just meant more conversations and more discussions. The reality is that if you’ve structured things well, you have layers of leadership. Not necessarily rigid hierarchies, but people across the organization who help manage it. I personally like staying approachable to anyone, but having those leaders in place is essential.
The real difference between a 10-person and a 100-person team isn’t just headcount. It’s about concentration of knowledge. With 100 people, you have more specialization, individuals who go deep on specific areas. With a team of 10 covering the same data and analytics scope, you often see the same person handling three to five different domains.
Because of that specialization at scale, you need excellent organizational practices. If those aren’t tight, the organization stalls and you don’t want to be the bottleneck.
Saket Saurabh:
Totally agree. Communication has to work because you’re aligning far more people toward a common goal. Looking forward, we’re seeing AI adoption actually shrink teams in some areas, reducing communication overhead. From the data and analytics side, has AI adoption changed how you think about team structure, size, and effectiveness?
Santiago Guillotti:
Before getting to that, on the communication point, it’s more than just communication. You need to establish processes that take decision-making off your plate when it shouldn’t depend on you. There’s a great book by Reed Hastings of Netflix, No Rules Rules, and I highly recommend it. You need to make sure the organization isn’t stopped by you. Communication is just one part of that.
On your question about AI and team structure, yes, roles are absolutely shifting. In teams I’ve built, I learned that a strong internship program can be a real asset. What I used to do was pair one or two interns with each senior member to handle tasks that were too small to automate but too operational for a senior person to own. The interns learned in the process. A win-win.
Now, a lot of those tasks can be automated with much less investment. What I’m seeing is that interns can actually work on building those automations themselves. And in many cases, they’re more comfortable with basic AI tools than the senior team.
The other change: previously, for every ETL job or data pipeline, you needed a certain number of data engineers writing all that code. That’s no longer true. Productivity is taking a big jump. And because of that, the demand for people who understand data and code is actually increasing, not decreasing. Data engineers are going to be more in demand, not less.
Saket Saurabh:
I agree, demand on data engineering has been growing. Greater AI adoption creates more demand for data from different sources. Can you give us more color on the real challenges of data itself? There’s a misconception that AI will automate everything, but clearly that’s not happening on the ground.
Santiago Guillotti:
People in roles like mine regularly get requests from business stakeholders saying, ‘My 17-year-old is doing this with Claude or ChatGPT. Why can’t we just connect all our spreadsheets and do the same thing?’
The challenge isn’t the AI component or the platform you put on your stack. The challenge is the legacy of the data, how your operational pipelines have been built over decades, and how you modernize that. Large organizations typically have thousands of databases, on-prem, often duplicated, often solving the same problem in different ways. It’s a real mess.
The problem isn’t adopting AI. The problem is data. You need to fix your data problems before you can properly connect an AI platform. And no one can fix your data problems for you. Vendors can help with your AI initiatives, but not with your data.
Saket Saurabh:
Absolutely. Without a good handle on data, trying to build higher-level AI applications is counterproductive. I’ve seen upwards of 10,000, and in one case 100,000 databases within a single large organization. Let’s shift gears to the insurance side. How is AI fundamentally changing the insurance sector?
Santiago Guillotti:
I’d say there are three main blocks.
The first is data handling and visualization across the organization. Today, you don’t need specialists for the visualization piece. Within two to three years, anyone who understands what they want to see will be able to prompt a tool and create that visualization.
The second block is the data and analytics team itself. Teams that used to receive few AI-related requests are now flooded with them. That’s driving operational changes across the business, in claims processing, policy management, and more.
The third block, and this is the core, is underwriting. AI gives you the ability to create micro-segmentation and risk-scoring algorithms that are far more precise. You can price risk much more accurately. And that ultimately increases insurance penetration across the income pyramid, in the US, in Brazil, anywhere.
Saket Saurabh:
So insurance itself as a product can become hyper-personalized. We’re already seeing that in automotive. Real-time data from vehicles is already being used for pricing.
Santiago Guillotti:
Exactly. It’s not that companies weren’t doing this before, they were. But previously it required enormous effort to build and maintain those segmentation models. Now it’s fast. Really fast.
Saket Saurabh:
As you automate the operational processes, you can actually service all those segments. One more area: large consumer-facing companies have been adopting conversational AI for customer support. What has made that viable, and how have you approached it?
Santiago Guillotti:
The first thing, and it may sound boring, is the security and legal framework of the country where you’re operating. In some places you can deploy conversational AI freely; in others, you need explicit disclosure.
But beyond that, the trick isn’t just replacing a telemarketer with an AI voice. You have to think about the entire process and automate the full flow. If a customer is filing a claim, you want the entire flow automated, not just the conversation. That means connecting all your legacy systems and data pipelines properly.
The AI voice tools are available. That’s not the hard part. The hard part is connecting your legacy systems in a way that enables full end-to-end automation. If you only automate the voice layer, the ROI isn’t there. But if you automate the whole flow, your clients are happier and you can likely break even within 12 months.
Saket Saurabh:
That makes sense. It’s the integration, not the AI layer itself, that’s the real challenge. For leaders at large, regulated companies trying to adopt AI, what advice would you give?
Santiago Guillotti:
The most important thing for any successful initiative is the business case. In regulated industries, it’s the business case plus the regulatory framework. The playing field is well-defined in terms of what you can and cannot do, so you always need both.
That said, in my experience, most of these projects die before they start because the company is too slow to make decisions and too many people are involved. That usually happens because there’s no proof of concept before the project goes through formal approval. My advice: build the POC first, before you even start the formal process. That alone eliminates 99% of the friction.
Saket Saurabh:
Looking ahead, what does a data and analytics organization look like five years from now?
Santiago Guillotti:
There’s a concept Peter Lynch talks about in one of his books. He calls it ‘diworsification.’ When great companies start earning a lot of money, they don’t know what to do with it and start diversifying into areas that aren’t their core focus. I think the same thing is happening in large enterprises with AI.
Companies that are making good money say, ‘We need to be in AI,’ and start building internal AI labs from scratch. But that’s not their core business, and it’s going to be a problem. They’ll burn resources building something they’re not positioned to win at.
Other companies will outsource the AI capability to specialists and focus their own energy on their data, because the data is their core business. The data you generate as a company is part of your value creation from day one. No one can give you that data. Putting an AI platform on top of that data is something any vendor can help with, but no one can replace your data. The winners will be the companies that make that shift.
Saket Saurabh:
That’s music to my ears as founder of Nikola, a company focused on solving enterprise data challenges. It’s not about building the next AI model. It’s about managing your data well and recognizing that data itself is the primary driver of competitive advantage.
You mentioned Peter Lynch on diworsification and Reed Hastings on No Rules Rules. Any other book recommendations?
Santiago Guillotti:
One more, it’s fairly technical but accessible even if you’re not deeply technical: AI Engineering by Chip Huyen. It’s a great read for anyone who wants to understand how these things actually work, specifically what we mean when we talk about an AI platform, and the distinction between the data layer and the AI layer.
Saket Saurabh:
Great recommendation, a solid reference for our audience. We covered a lot of ground today: technology, team building, the data challenges behind AI adoption, and the insurance sector. Thank you so much, Santiago. You really brought the conversation back to what matters most, data and its unique role as a competitive asset. Thank you for being here.
Santiago Guillotti:
Thank you, Saket. That was great. Let’s talk again soon.
Saket Saurabh:
Absolutely. Thank you.
