How to build data foundations for AI exploration

The critical importance of data is nothing new to CIOs. Digital leaders spend years collecting and collating enterprise information. This data is squirrelled away in in-house data centers and cloud-based infrastructures with the hope that one day it’ll be useful. Now, with the rise of AI, that time has come.

CIO.com’s 2025 State of the CIO Survey found that researching and implementing data-enabled AI products and projects is the top CEO priority for IT. Bosses want their CIOs to move beyond experimentation and deliver value in collaboration with business leaders, whether through productivity boosts or efficiency gains.

A similar story emerges from the Carruthers and Jackson annual Data Maturity Index, which found that over 50% of organizations have increased their use of AI in the past year, with 11% reporting a significant increase. Caroline Carruthers, CEO at the consultancy, says the focus on AI has a commensurate impact on interest in data.

“What we’re finding is there’s almost been a second dawn of realization that data is important,” she says. “It’s the same kind of effect we saw with the introduction of GDPR in 2018. I think that was probably the first dawn and the AI effect is the second.”

While organizations are eager to explore AI, they’ll only be able to exploit emerging technology if they have reliable and scalable data foundations in place. To this end, the cloud-based infrastructures CIOs have built over the past decade will play a key role helping their companies scale IT resources in response to new business requirements.

However, AI ramps up the pace of change. In an era of almost constant transformation, public or private cloud provision is simply one piece of the IT infrastructure puzzle. Howard Holton, COO and CTO at researcher GigaOm, suggests building effective foundations for AI requires a nuanced understanding of infrastructure that many business leaders lack.

Many organizations shifted wholesale from traditional in-house storage to cloud-based services, a broad-brush approach that’s imperfect for the fluctuating demands of emerging technologies. Holton says the better approach is to build your data foundations for change and a future that’s likely to involve a significant shift in direction every 12 months.

“There’s a tremendous amount of churn in the space, and the market is evolving too fast to make hard decisions,” he says. “Digital leaders must design their architecture with the knowledge they’ll be wrong. The right approach is like how we architect for cybersecurity. It will fail, so build that knowledge in so you can change pieces as needed.”

Making data the platform for everything

One digital leader who recognizes the critical role of agile foundations is HPE global CIO Rom Kosla. His company uses third-party applications with built-in AI capabilities, including Microsoft Copilot and tools from SAP and Salesforce. HPE has also spent 18 months developing an internal chat solution called ChatHPE, a gen AI hub for internal processes powered by Azure and OpenAI technology.

Kosla says new business proposals for ChatHPE are assessed to create a pipeline of use cases, such as reviewing legal contracts, boosting customer service, reusing marketing assets, and improving financial analytics. The pipeline is managed across an IT infrastructure that includes HPE Private Cloud AI, a full-stack solution for enterprise AI challenges. Kosla adds this platform allows his organization to scale data resources for AI projects effectively and securely.

“The IT department was one of the first to receive the Private Cloud AI servers and put them into our collocated data center,” he says. “And now we have a routing mechanism that says if we want specific internal AI use cases that we don’t ever want to be trained and inferenced on models outside, we can keep that within that data center.”

Understanding the use case for data is also crucial for Steve Riley, head of IT operations and service management at Mercedes-AMG Petronas F1 Team. His team consumes huge amounts of information to improve on-track racing performance, and a great data platform for his IT department is scalable and, more importantly, ultra-reliable.

“We do a lot of computational fluid dynamics, so having a high-performing storage layer that sits underneath is of high value to us,” he says. “But success is also about having the right technology in the right place. And that doesn’t always mean just buying the quickest, fastest, and most expensive technology. It means being strategic in terms of where you place your investments, and that’s the approach we’re taking.”

Riley says Mercedes F1’s main technology partners include HPE for data infrastructure and TeamViewer for emerging technologies, including its Tensor platform for race simulators. The team is also dabbling in AI. The right platform, says Riley, means his organization can make data-enabled maneuvers at the appropriate moment.

“It’s our job to apply our industry knowledge,” he says. “That often means we need to think long and hard about where we spend our money. We’ll look for performance where it’s needed. However, reliability is key because then the team isn’t continually going back over the same ground again, and they’re freed up to look at the next big thing.”

Richard Masters, VP of data and AI at Virgin Atlantic, is another digital leader on the lookout for innovative solutions to intractable challenges. A key element of this approach is the organization’s Databricks data platform. His team uses the platform to consolidate enterprise information, and consider how to react to new technologies and models that emerge.

Masters says easy access to information has a big impact. “When a question came up in the past, when we didn’t have this insight, our team had to go into a SQL Server database over here, or a Postgres one over there, or an Oracle one, or then go and talk to the system owner and then remodel the data,” he says.

Virgin’s Databricks platform means enterprise information is now stored in one location. Masters says his data team can spend more time exploring, understanding, and validating information. That capability means the team can unlock insight much more quickly as questions from the rest of the business come in.

“That’s a massive gain because data is the foundation of everything,” he says. “This capability means you can do more with emerging tools like AI agents because you have trust in the data in the platform, and you can reduce a lot of the noise.”

Staying open to innovative ideas

Digital leaders, therefore, must recognize that ordered data is crucial to exploiting innovation. As Snowflake co-founder and president Benoît Dageville suggests, CIOs should think of enterprise information as “layer zero” of the data foundation. CIOs who focus on this groundwork now will be well-placed to exploit emerging technologies later.

“AI is driving innovation faster than anything we’ve seen,” says Whit Walters, field CTO for cloud and data infrastructure at GigaOm. “Most tools were designed a few years ago to manage the market forces driving cloud. Everyone is just trying to keep up. In the meantime, businesses need a modular architecture. The parts of your data platform must be interchangeable as vendor developments in the age of AI will be asymmetrical.”

That’s an approach that chimes with Vivek Bharadwaj, CIO at clothing manufacturer Happy Socks. Before migrating to a Snowflake data platform, his business was dependent on Excel for reporting. His team built greenfield data foundations using Snowflake to help the rest of the organization make quicker, insight-enabled decisions. They continue to add new technologies to build out the platform.

“At first, we were running purely on Snowflake,” says Bharadwaj. “As we got bigger, we migrated more into DBT and a partner ecosystem stack that includes Airbyte and Sigma. As our process matures, we’ve started exploring different apps, like the Streamlit framework and Cortex AI, to deliver business value.”

Happy Socks also uses its data ecosystem to explore new use cases. Bharadwaj says potential areas of investment include predictive analytics, inventory management, and personalization for e-commerce customers. Plus, his team is investigating how Snowflake Cortex LLMs can be used to create product descriptions, freeing up employees to focus on higher-value issues.

“The real goal is to use data and AI to solve our business issues at scale,” he says. “The thing I’m a big proponent of, and we’re making meaningful strides in this area, is that if you want to grow non-linearly with data within your company, make it self-service from day one. The business users should be empowered, and they shouldn’t depend on IT for every support issue.”

And in terms of lessons learned, his recommendation to other CIOs is to start with the foundations and don’t build AI on shaky data infrastructure. “You may get some initial wins, but it’s never a sustainable approach,” he says.