Snowflake Summit 2024: Why IT Leaders Must Reimagine Data Governance

The more businesses leverage generative artificial intelligence, the more robust their data governance must be. At this year’s Snowflake Summit 2024, industry experts discussed the right way to approach data, and why the governance model — typically considered a strict process of managing the security and integrity of data across the enterprise — must be reimagined for the future.

“Historically, we’ve seen data governance as something to lock down and control our information from a security lens. But in the era of AI, we should be looking at data governance with a new lens, and that is around how do we make our data more usable? What can data governance help us do beyond security control and how can it help us get usability out of our data in the context of AI?” said Wendi O’Neill, senior director of the data and analytics presales team at CDW.

Seeing data governance as more than just a form of secure quality control and instead as a prerequisite step that allows businesses to transform workflows with generative AI is a major shift.

“I think that's where many data governance programs fall short. We haven’t made that cultural shift to say data governance is actually going to make data more accessible and more usable for organizations,” she said.

“Everyone realizes that you need to have a good set of governance in place,” said Paul Zajdel, vice president and GM of data analytics at CDW, but there’s still a reluctance to embrace it, as though it’s an academic concept. “I come back to companies that still have a legacy mainframe or on-premises storage, and they want to skip way ahead and implement AI or build a modern data platform. You need data governance in place to do any of it,” he said.

This is precisely why data governance needs to be redefined as the gateway that makes enterprise innovations possible. Here is what IT leaders need to know to set themselves up for success:

For IT leaders across the country, governance is a top priority for data initiatives, and it’s one that they will need to invest in to make any new AI initiatives worthwhile.

Between now and 2025, 80 percent of organizations seeking to scale digital business will fail because they do not take a modern approach to data and analytics governance, notes a Gartner survey.

Data governance is crucial for generative AI projects because AI models must be trained on clean, trusted data. “When you're moving to cloud-based solutions, they should be able to handle these larger data sets and provide that computing power you need for your machine learning models,” O’Neill said.

Once high-quality data is in the mix, IT leaders can go one step further and integrate machine learning and AI into their data pipelines to improve predictive insights and optimize operational efficiencies. Snowflake’s unified platform, integrated with data science and ML platforms such as Amazon SageMaker and Azure Machine Learning, allows teams to build, train, and deploy ML models directly within the Snowflake environment, for example.

Once businesses establish data governance frameworks, IT leaders can leverage those data management strategies toward high-yield AI projects.

“All roads point back to governance,” Zajdel said. “We can start anywhere with our customers, wherever they are in the journey. But no matter where they want to go, they need a good foundation of data governance, which is really data quality.”

With well-governed data, AI can analyze customer data to identify user preferences, forecast demand in supply chains and safeguard financial transactions. Teams can also embed data governance rules into AI algorithms to ensure adherence to set privacy policies or run compliance audits.

“The cleaner and more high-quality your data, the better,” O’Neill said. This step makes it easier to “integrate your data from all different sources into AI tools.” It can also help synchronize data across workflows and achieve compatibility with legacy systems, which is particularly important, she noted.

Industry experts at the Snowflake Summit also shared how adopting Snowflake’s scalable data mesh approach strengthened their data governance further, because the platform allowed for decentralized data ownership, role-based access control and end-to-end encryption. The result was that teams could track the progress of any given project and ensure that each of their AI investments delivered tangible value.

Intrinsically linked with large language models and data governance is the issue of ethics. Just this year, studies highlighted a 600% increase in enterprise AI adoption, which helps communicate the scale of data being used — and the potential consequences, if left unchecked.

“We have to really consider our ethical use, not just for bias, which is hugely important, but also to understand that if we’re making decisions based on AI, they better be accurate,” O’Neill said.

With management and use policies in place, businesses are much more protected and accountable.

For O’Neill, data governance also offers businesses a rare “lineage” for the data itself. If an issue arises such as an AI hallucination, teams can trace its origins.

“It’s important to understand where the actual information came from, because with all the regulations that businesses face, IT leaders need a proof point.” Businesses need “to know where that data comes from and whether it’s accurate and true.”