Kyndryl CTO: Our AI Mission Is To ‘Help Our Clients Become More Intelligent’

‘There’s such a renewed appetite [from clients] to truly understand how their business operates. And this is the first time, with tools like AI, like GenAI, that we can start to build in their view, their context,’ Kyndryl CTO Antoine Shagoury tells CRN.

Antoine Shagoury, CTO of solution provider giant Kyndryl, is seeing growing demand from customers in the generative artificial intelligence era for partners who can deliver the best AI-driven insight, eventually moving into predictive and prescriptive intelligence.

“There’s such a renewed appetite to truly understand how their business operates,” Shagoury told CRN in an interview. “And this is the first time, with tools like AI, like GenAI, that we can start to build in their view, their context. You’ll see tremendous opportunity happening in that.”

The CTO of the New York-based company, No. 10 on CRN’s 2024 Solution Provider 500, shared his insight as developments including the $500 billion Stargate AI infrastructure project by OpenAI, Oracle and others plus DeepSeek’s apparent AI model cost-efficiency breakthrough help make the emerging technology a greater possibility for organizations to adopt at scale.

[RELATED: Kyndryl Shoots To Q3 Profit With ‘Strong’ Signings Growth]

Kyndryl AI

Shagoury said that emerging businesses within AI—including agentic AI and private AI—are opening new opportunities for the solution provider.

He estimated that about 20 percent of Kyndryl’s customers have introduced AI into operations. More than half are trying out AI.

“It’s probably doubled in the last year,” he said. “I could probably say it’s going to triple over this coming year.”

Embracing the AI era doesn’t mean customers are turning their backs on mainframes and legacy technologies, said Shagoury, whose company spun off from mainframe giant IBM in 2021.

“The amount of advancement and integration capabilities that we’re also seeing in the upgraded processors within the mainframe, it’s keeping pace,” he said. “It’s embedding the AI capabilities. It’s helping to provide opportunities to improve operations, MIPS [millions of instructions per second] utilization, cost. It’s making the mainframe more flexible, even for data on-boarding, off-boarding. Things that traditionally had a lot of customers ready to move.”

Here’s more of what the solution provider’s CTO had to say about Kyndryl’s AI opportunity.

How is the AI era affecting Kyndryl’s business?

[Implementation and getting into a business’] operations–that’s where I think the real opportunity is for us.

It’s not only just incubating—how we go from ideation to demonstrating business value—now we can take a factory approach. How do you want to try and take advantage of these new technologies and capabilities?

Even just the barrier to cost, getting into this. Investments like [the $500 billion Stargate Project investing in AI infrastructure for Microsoft-backed ChatGPT creator OpenAI] can start really opening up opportunities.

There is a good upsurge in private AI type facilities routing workloads—so properly balancing AI workloads with the most cost-effective models. Sounds a little bit like cloud.

We’re definitely getting into a good reflective phase on how the companies have been approaching it. What we are now doing is seeing where there is opportunity, or how that can now scale and be adopted.

Can we actually ride the opportunity curve with [Stargate]? Are we going to see economies of scale? Are we going to see faster time to markets and more industry specialization? And I’m an optimist here. I think the answer is yes across the board.

Are customers putting AI into production?

Some stuff is doing far better than others. If we’re looking at automotive, there’s already been decades of investment in robotics and intelligence around facilities, supply chain.

There’s an easier path of how we get into looking at quality, agentic opportunities. They’re further down the curve and the path.

Some pharmaceutical areas are definitely further ahead than what I call traditional business. There are some spaces where we’re seeing … a lot more interaction on travel. Agent listeners. How do I improve your experience?

Other companies are still in the POC [proof of concept] phase. Maybe 20 percent of the clients have AI in some form of introduction in operations and business.

I have two dimensions of using AI. One is embedded, where you may see partners like us. We use it. How is that introduced to the customer in regard to operations insight. The ServiceNows, the Salesforces.

Those are just what I call natural on-boards. And then there are the true infusions. And that’s where it’s a very small community.

More than half the clients are trying something somewhere. Whether it’s copilots, whether it is around operations, whether it is around analytical harvesting and looking for business patterns. But that’s still in the early stages.

It’s probably doubled in the last year, and I could probably say it’s going to triple over this coming year.

How often are CEOs talking about AI? If you look back two years ago, it was almost nonexistent. Last year, it was in like 80 percent of the announcements and the publications and the earnings releases and the strategy documents.

There’s a change, actually, in willingness to try and adopt it. That’s what I think means a lot to Kyndryl and the market that we serve.

Not only three, five years out. We’ll be a strong player that far out, too. But we’re one of the well-positioned players to help them in the interim over the next one to two years on really grounding out their strategies.

What obstacles remain for customers to adopt AI?

Cost efficiency is still out there.

It’s usually in the first sentence. Does it help me find more efficiencies quicker?

The new one that’s growing … is effectiveness. And that gets into regulation, compliance. Can I be more effective in my business operations?

And we’re seeing a good uptick in highly regulated businesses and sectors.

They’re looking at it as a boost or an accelerant to their positioning in that. There’s some good space there. And then that’s when we kind of get into what I call the ‘experience change.’

Travel industries, even automotive. How can I change the experience of the buyer, the user?

Is the AI era affecting data centers?

There’s still significant data center presence. It’s probably the area that has the most consistent amount of scrutiny on further investment.

A lot of colocation investment. A lot of migrations. And we run them too, by the way. So even our view of physical data center facilities has really been put under a different type of scrutiny.

So now we look at higher-density, higher-compute capabilities, hyperconverged capability. So it’s not like we’re getting away from them.

The question is, can they meet the financial demands? Can they meet the power demands? Compute demands?

There’s a renewed look at how can you invest and take part in these compute facilities. So kind of that cloud mentality again.

The more recent discussions are, on the one end, trying to mature the cloud-type adoptions. And how does my business become more portable, more fluent in operations?

And now with this, how do I look at taking advantage of this temporal compute, this capacity to really look at analysis in my business differently? There’s a convergence there.

Data center is definitely not dead. They are going to evolve as fast as we can possibly continue to push them.

Compute capabilities are feeding that wonderfully. You’ll still see the shift from owned facilities to quality-of-service-type of engagements with physical environments like that.

What are you seeing with private AI?

We’re investing in it. We’re partnering in it. We have a partnership with Nvidia.

[We had a] recent announcement with Dell and Nvidia in Asia and Japan, specifically on building these sovereign-type environments that help address the concerns on data, on data risk, on the exposure of the analysis themselves and even how we build into the life cycle protecting information, think even data shredding types of concepts.

How do I not only provide the facilities, but give the certainty that the information, the analysis and the capabilities in the IP are protected?

There will be continued investment in that. You’ll see a lot of growth. It’s a more comfortable space for many businesses to go to.

How is Kyndryl growing its AI capabilities in 2025?

You’ll see a continued rise … on building more intelligence in the services in which we provide to the customer and what we do with our partners. Can we continue to differentiate?

There’s a huge investment pipeline on how we continue to integrate across partners. We want to make sure our client gets the best.

They want the output. They don’t want to deal with the mechanics, the integrations, things of that nature. So the more we can develop into that space, that is absolutely an enabler and a boost to clients’ businesses, especially in how they apply their investment dollars to their business. There is huge growth in that space.

I’m a big advocate of constantly echoing back how we can help our clients become more intelligent. So on top of how we provide the tooling to do that in how we provide our services, can we help them through where they can build better insight into how the business operates?

There’s such a renewed appetite to truly understand how their business operates. And this is the first time, with tools like AI, like GenAI, that we can start to build in their view, their context.

You’ll see tremendous opportunity happening in that.

What we’re investing a lot of time and skill to bring forward to customers is in how do I go from simple automation, simple AI application, to really build the dataset to drive into patterns?

Once I get into predictives, I can then get really into prescriptives. And that’s where the agent overlay becomes very much a boost and accelerator to demonstrating how it can help their business.

Financial service institutions looking at transactions, transaction errors. How do we catch them faster? How do we go back, make sure it doesn’t impact the operating performance of the business or a client account?

Will AI PCs help grow adoption?

There is an intelligence everywhere dilemma.

If I’m on a single application on my phone working on one thing, having it localized makes a ton of sense.

How do I extract that from an individual to a business benefit, to a service benefit, a producer benefit—that’s where there needs to be tremendous opportunity to integrate the insight, to build these insight engines.

There’s a lot of early discussion on what is the evolution of inference engines? What is the evolution of data integrations or insight integrations that can really look at the business in the way it’s intended?

We’re in an arms race, so to speak. Microsoft wants us to focus on Microsoft. Amazon wants to focus on Amazon insights. Google, their own.

Look at language models. There’s such a competition over the language models and the small language models and micro language models.

If I look at all the parameters across the models, there’s over 3 trillion to almost 4 trillion in the parameters. Significant overlap. Imagine if the insight, the context, were able to take advantage of that entire population. That’s where the amplified benefit comes from.

Where is the AI era going in the long run?

There’s a ton of history that could be a great example as to what can easily manifest here.

[AI] chatbots literally go at it with each other. Have a comment, have a response. They can go on forever.

And I spent most of my career in financial services. Trading, for example. High frequency trading. Programmatic trading.

When we had the largest issues, we had programs fighting programs. Pricing systems fighting pricing systems. You could see that it’s the war of the bots, the war of the AI.

You can easily escalate without the right balance. And then the financial services market … some of the crash issues and some of the price issues, some of the companies, a lot of these programmatic interactions have disastrous impact.

You always learn from the lessons, but this is the same type of situation.

The market benefits by the growth and the investment. The rate of evolution of business right now is unprecedented. It’s the greatest we’ve ever seen in our lifetime.

We’re the good stewards. We benefit, one, by our positioning as probably one of the largest and the most stable operators of mission-critical systems.

Can we integrate it in a way in which that builds confidence? Yes. Can we integrate in a way in which that builds transparency? Yes.

That’s where we win and we help the client win.

We all rise and sink with the tide. As the markets improve, as the businesses improve, adoption improves, we are the natural beneficiary.

What outside AI is fueling demand for Kyndryl?

We [the channel and technology executives] spent the last 10-plus years on … how do I get to cloud, hybrid cloud, private cloud solutions? Application implications.

We didn’t really solve a lot of the technical debt, but we made a lot of complexity in building out these straddling organizations.

Then we got very deep into SaaS opportunities. Went from a few 100 partners to a few 1,000 SaaS providers. Even further complicated those challenges.

And then we got into the era of observability but without integration—12 to 30 observability tools.

You create a whole new mess in that modernization concept. The things that have been driving tremendous benefit to us and with our customers is we’re now building the agnostic views of observability. And it demystifies what we do, how the client is operating.

You’d be surprised at how many clients are surprised in their own business ops of how things are running.

It’s shifting to not just platforms within a company, but can we build ecosystem … capabilities that the clients can start to simplify, get better observability into how their business operates, and almost immediate improvements in how they operate.

That, to me, is almost a predicate to when we start looking at automation, agents, further enhancement.

How is Kyndryl leveraging AI internally?

When we first became independent [from IBM] … we were doing about between 10 [million] and 20 million automations a month.

Now we’re at about 200 million automations a month. A lot of it is orchestrated through different AI triggers and models.

Now I can get into better predictions. Now we have literally tens of millions of intelligent insights that have now taken account and look at different AI models for patterns, pattern avoidance, issue avoidance, early alerting.

As we’re now starting to get deeper into that, this is where we’re now deploying agents in our clients. So an agent can now not only help develop part of the code, it can actually be a part of testing the code. Orchestration agents can move information around, can segment queries. You get faster opportunities to diagnose a problem.

When you get into, say, manufacturing and see some of the automotives, you can not only look at the reference information … but we can now have the segmentation look at, hey, what are the patterns of repairs that have happened in the past?

We can now reduce the time in which they take to diagnose a new problem, an issue or resolve it. Those things are happening now. We’ve seen an amazing uptick in what I call our ‘embedding of AI,’ of AI, tooling, models, analytics and now agentic opportunities in what we deliver to our customer.

What happens to mainframes in the AI era?

The amount of advancement and integration capabilities that we’re also seeing in the upgraded processors within the mainframe, it’s keeping pace.

It’s embedding the AI capabilities. It’s helping to provide opportunities to improve operations, MIPS utilization, cost.

It’s making the mainframe more flexible, even for data on-boarding, off-boarding. Things that traditionally had a lot of customers ready to move.

It’s a proven horse, and it’s still running well, right in that space. But we’re agnostic. If a client wants to move, our teams use the tools to look at what we can modernize the code with. We can update the code. We can port.

How do we move the right workloads on the mainframe? And how do I get the right insight to evolve it and maintain it? We still have a strong commitment to invest and develop and advance the mainframe. And our goal is really not on the what, but the how.