How Kyndryl’s separation from IBM prepared it for the AI era

In 2020, IBM announced it was spinning out its managed infrastructure services division as an entirely independent business. It would require essentially building a company from the ground up with everything from a new network to a new security posture to physical badges to get into the building.

Yet it wasn't really starting from scratch. Kyndryl was the size of a full-blown company, and it had data and applications and a stack of legacy technology that it had built up over time as part of IBM. The spinout gave the company the opportunity to put its tech house in order, to pay down its tech debt and to build a modern foundation for the future.

At a time when companies looking to shift to AI are dealing with a monumental change management problem, Kyndryl CIO Kim Basile and her team have already been through one. Along the way, as the company separated from IBM, it pared down its application pile, chose modern systems, and without even knowing it, prepared itself for the AI age.

I spoke to Basile about her company's journey to independence, the challenges she faced along the way, and the ways in which the act of going on its own forced a step change that may otherwise never have happened.

Two year transformation window

Although the spinoff was announced in 2020, the two-year clock to become fully independent started ticking as soon as Kyndryl separated from IBM in November 2021. Basile actually joined the company in May 2022 when the process was well along, but she understood how difficult it was to pull off a transformation of this scale and complexity in such a short period of time.

A normal transformation process starts with figuring out your as-is state, then looking at where you want to go and mapping out a way to get there. This process  can take years and they didn't have that kind of time.

"We literally had two years to rip out a network, stand up all new applications, get new strategic tools and stand up a security system. We didn't have time to do [any of the normal up-front work], so we had to drive simplification by default into what was out-of-the-box functionality of the tools," Basile told FastForward.

They also used that time to analyze their inventory of applications that had built up over years and take a hard look at what they were taking with them to the new company, and what they were abandoning by the side of the road. In the end, they took a library of over 1800 applications and winnowed it down to around 360, according to Basile. 

"As you can imagine, those 1800 apps all had parents that loved them and didn't want to give them up," she said. It took some tough decisions, and some not-so tough ones. Of the more than 1400 applications they ended up dumping, over 400 were Lotus Notes apps, and every last one of them was put out to pasture.

Prepping for AI

A lot of what they were doing from updating apps to getting their data house in order were painful things every company trying to get their company ready for AI is doing right now. It was a happy coincidence that the company was prepared for the modern AI era, and it didn't escape them. "It gave us a foundation and many companies are still trying to build that foundation," she said. 

That foundation included building a proper data catalog that helped it understand where the data resides across the organization and where the gold (authoritative) copies are stored. The network itself was modernized to roughly 99.9% on SaaS and cloud, moving on from the old IBM on-prem infrastructure, according to Basile. And when AI arrived, the company started with governance before any rollout. "The very first thing we did was stand up an AI governance council," she said.

Beyond the technical aspects of updating all of their technology systems, the exercise of trying to figure out how to get this done had an even more profound outcome. "I would say, honestly, about what it got us ready for was quite candidly was to be able to think differently. And in my mind, AI, you have to think differently," she said. And that means giving up a lot of preconceived notions about how you do things.

That's something I'm constantly hearing from execs about building agents. You have to reinvent processes, rather than slapping AI on an inferior way of doing things, and Basile says that aspect of reviewing processes is never really done. "It doesn't mean we, in some cases, maybe didn't drift back a little bit. And so as we're adopting AI, we've got to streamline those processes again," she said.

She said the mission to get independent so quickly was really a once-in-a-lifetime opportunity and really helped them modernize and get ready for AI, but she cautions that employees can get overloaded with too much change.

"You have to make sure that your organization doesn't get fatigued from constantly transforming, but embraces it in a way that get employees excited about it. They trust it. They want to learn from it. They feel part of that journey."