AI may dominate the headlines, but at its core, data integration remains a deeply human challenge. This was one of the standout themes from our recent conversation with Markus Kolic, Associate Director of Engineering at Sun Life U.S., on the latest episode of our Behind the Data podcast.
Markus shared his insights on making progress in the rapidly changing world of AI, the complexities of integrating legacy systems with modern platforms, and the importance of understanding the human factors behind data systems.
”All of these things are people from the beginning,” said Markus. “All of the systems that we build, all of the software that we have, is here to serve the people."
Here are some of the key takeaways from our discussion.
Experimenting with AI in a world of unpredictability
We’re living in what Markus describes as ‘a moment of incredible possibility’ for AI. Yet, while large language models like ChatGPT are generating immense hype, Markus cautions against rushing to implement AI-driven solutions without fully understanding their potential and their unpredictability.
“The basic challenge in a business that is regulated and profit-driven,” Markus explains, “is that generative AI by definition involves some unpredictability. It’s a non-deterministic system. You could put the same input into it twice and get two different results.”
This unpredictability can be unsettling for enterprises that thrive on consistency and control. Markus stresses that organizations must allow room for experimentation, failure, and rapid iteration if they want to harness AI effectively.
“You need room for experimentation. You need room for iteration. And you need room for failure,” he says. “The way to succeed with AI is going to be to take a few smart people and put them in a room and let them experiment for a year.”
A practical approach to legacy systems
For organizations dealing with legacy systems, Markus advocates for the strangler pattern: a strategy that allows businesses to modernize incrementally, rather than attempting a risky, large-scale replacement.
“The strangler pattern is a model for moving [legacy systems] gradually,” he explains. It is inspired by Martin Fowler’s article on the strangler fig. “Instead of replacing your whole system outright, you go piece by piece. You rewrite one portion in a microservice or in whatever your new system is, and you link them together using APIs.”
This incremental approach helps minimize disruption and allows teams to adapt as they go. However, Markus warns of the risks associated with getting stuck midway.
“If you stop halfway through… you’ve made the problem worse instead of better,” he says. “Inherently, when you’re following this kind of pattern, you are making compromises and temporary workarounds along the way. And there’s the old truism from corporate software development: There is nothing so permanent as a temporary solution.”
Human-first data mapping
Sun Life’s ongoing efforts to integrate and modernize its data systems reflect the importance of a human-centred approach to data mapping. Markus describes his team’s process of creating a domain-driven data model that reflects a shared understanding of key concepts like insurance policies and member records.
“Data is not information until it can be used and understood by a human,” Markus explains. “You need to understand the people that it is from, the people that are handling it, and the people that it is for.”
This process, however, is far from straightforward. Markus likens it to archaeology:
“Older systems… embed their own assumptions about what that data means and how it works. There’s archaeology involved in all of this as well. You’re trying to figure out what those concepts were in the first place.”
To ensure that this work is collaborative and adaptable, Markus’s team has moved its data mappings to GitHub, using YAML files to create a versioned, living document.
“Your data mapping needs to be a living document,” he says. “It needs to be something that can be collaborated on and shared by everybody… We store these mappings in readable YAML, so when anybody working with our teams wants to update this, you edit the YAML and open a pull request and we can see exactly what is changed and when and by whom.”
A flexible partner for integration
CloverDX has been central to Markus’s work, serving as a flexible, scalable platform for data integration at Sun Life. One of its standout benefits, Markus says, is its compatibility with agile development practices.
“Our team using Clover as its platform… can get a request for some new piece of data, and build it almost immediately in Clover. Just drag a couple of components onto a screen, define it, open a pull request, get it reviewed, merge it, test it, deploy it in the space of maybe a couple of hours. In a corporate context, that is astounding.”
This agility has allowed Sun Life to scale its data operations while maintaining flexibility, making CloverDX a powerful tool for tackling both legacy and modern data challenges.
By emphasizing the human element, organizations can ensure that their data initiatives succeed and deliver meaningful value.
To hear our entire conversation with Markus Kolic, check out the full episode.
If you’re ready to see how CloverDX can transform your data operations, get in touch with our team today.
Share
