NVIDIA GTC 2024: Is Your Data Optimized for AI?

Generative artificial intelligence is a major point of discussion for CEOs and boards of directors as IT leaders look to take advantage of this transformative technology.

“Organizations are not just talking about generative AI. They’re investing time, money and resources to move it forward and drive business outcomes,” noted Gartner Distinguished Vice President Analyst Frances Karamouzis in a press release on a 2024 Gartner poll of executive leaders. “In fact, 55% of organizations reported increasing investment in generative AI since it surged into the public domain ten months ago.”

But before companies get too far along in their AI plans, they must organize their data ecosystems. CDW experts Rex Washburn, chief architect – FCTO data, and Wendi O’Neill, senior director of data analytics and presales, said that this foundational step is vital, and that many IT leaders are mixing up the necessary order of operations. Without a properly structured data environment, even the most sophisticated AI technologies can falter, leading to inefficient operations, skewed analytics and missed opportunities.

At global AI conference NVIDIA GTC 2024, hosted in San Jose, Calif., experts explained that organizing data ecosystems not only enhances the performance of AI applications but also ensures that companies can fully leverage their data assets for more informed decision-making.

Here are a few steps that organizations should take to prepare their data ahead of AI implementation:

There’s an enormous appetite for AI right now, but many companies have not yet done the work to prepare their data systems properly, Washburn said.  AI systems require high-quality, accessible data to function effectively, and if organizations haven’t done the work to collect, centralize and clean their data, ROI will be limited.

To begin this process, IT leaders must embrace a new approach to data. “It’s really just a change in thinking. We’re doing a lot of the same things, but the architectural approaches that gave us these monoliths need to be left behind, because we need to move at the pace of the future, the pace of AI, and traditional platforms just really don’t do that,” Washburn said.

A future-oriented model is all about data fabric. Data fabric is a “unified platform that includes the integration, the security and governance of all data,” Washburn said. It brings a “diversity of data,” whether it’s stored in the multicloud or on-prem, into one accessible space so that teams can view, sort and extract insights from it.

This setup offers far more “interoperability, modularity and flexibility,” so teams can “rapidly onboard and also move products out of the system,” he said. The more scalable the architecture, the faster teams can innovate from a quality data source.

“Unifying the data” and getting it cleaned, sorted and managed should happen before AI is added into the mix, Washburn said. Establishing data pipelines well in advance will make AI and machine learning experiments much more successful.

IT leaders can also work with a variety of vendors to handle the virtualization or federation of data. CDW can also help at this stage: Experts can review a company’s IT infrastructure, run an analysis and curate a tailored roadmap to a data fabric setup.

A formal data governance program outlines how data for AI models will be used, managed and protected. It sets clear policies for data use, compliance and privacy, which are particularly important as AI applications can produce errors and reflect biases.

When data is private, accurate and secure, AI systems can generate more useful insights, and IT leaders can make data-driven decisions with confidence. “A program around data governance helps you know the quality and trustworthiness of your data,” O’Neill said.

By setting these standards ahead of time, organizations can mitigate risks, enhance the effectiveness of their AI systems, and ensure that their use of AI aligns with both regulatory requirements and ethical standards.

“Data governance drives literacy across your organization, not just within the business users but also within the folks who are managing the data,” O’Neill said.

Once data platforms, pipelines and governance programs are established, IT leaders must do the hard work of change management. This means making sure that teams use these tools every day and treat “data as a strategic asset,” Washburn said.

“When we work with our customers, it’s all about helping them be in a space where data informs everything they do. It’s part of the DNA of their business,” he said.

If IT leaders can engage directly with data from the start of any project, it will be easier to change direction and adapt to future demands. Creating a culture of data literacy is also about empowering employees to understand, interpret and question the data underpinning AI outputs. Cultivating a degree of skepticism and critical thinking can ultimately lead to better decision-making.

For example, if the data is fluctuating in real time, can a single source of truth be determined? And if not, “do we understand what could be true and what could be false?” O’Neill said. That follow up question is essential and requires that employees are comfortable interrogating the data.

That’s the intellectual exercise required to move the business forward. “From there, I can engineer from the back end for the best cost, the best performance, and that’s going to be key,” Washburn said.