An AI Adoption Imperative: Centralized Sources of Governed Truth
A Q&A with Cody Irwin
As the barriers to entry in AI dip lower and analytics are offered as self-service, data designers ponder the steps to making AI adoption succeed with centralized, governed data. Here, we speak with Cody Irwin, Domo's AI adoption director, to ask for his insights about adoption strategies for enterprise teams who aim to build a data foundation that will move the institution from AI experimentation to real-world execution.
Mary Grush: We've heard about the perils of siloed data for years. How does this change now, with AI?
Cody Irwin: Data access and governance hold some of the biggest hurdles to unlocking the promised efficiencies behind generative AI. Leaders have been told for years that building a data warehouse, or lake, is critical to visibility and decision making for analytics. That need has now become an imperative. The barrier to entry is no longer "Do you know SQL, data science, and visualization techniques?" It's simply, "Do you know words?" To empower the enterprise, data leaders must create centralized sources of governed truth.
Grush: Is AI well understood in the context of higher education data governance? Is there a "trust deficit" to overcome?
Irwin: The trust deficit exists everywhere but is especially acute in higher education. Education institutions rely on data to manage admissions, financial aid, research, publications, accreditations, fund raising, compliance, and operations. A misstatement of data is often public and can have serious ramifications on institutional credibility. Since data access and analytics are becoming ever more self-service, data leaders have a burden of responsibility to create and manage centralized, certified data.
Grush: I know these are complex issues, and our time is short here, but what are some of the qualities of data models that design leaders should be working towards?
Irwin: We've seen value in data modeling that has AI — and flexibility — in mind. Specifically, institutions should consider adopting a "medallion architecture" where "gold datasets" are exposed for use by decision makers. Also, AI thrives on context, which requires more than just making the data available — the data models should surface semantics that provide organizational context that AI can leverage to provide more meaningful and accurate responses.
Grush: Could you give an example of how designers can work effectively with data productivity platforms?
Irwin: Step one for most data designers is to make the data centrally available through a governed interface. They should provide the ability to retrieve or integrate data from almost any source environment. That centralization should allow for the implementation of policy, security, logging, and certification. The centralization should be empowering and not restricting for decision makers. If it's not easy, people tend to find a way around it. My company, Domo, provides controlled interfaces on top of that centralized data fabric for self-service analytics and easy AI interactions.
Grush: How can people be strong design leaders in a shifting AI culture with massive data needs?
Irwin: The data foundation is critical. The faster designers can get a foundation in place, the more quickly their internal customers will feel empowered. We recommend not letting perfection be the enemy of progress. Design leaders should prioritize what they feel would be the most impactful and move quickly to get that released.
[Editor's note: Image by AI. Microsoft Image Creator by Designer.]
About the Author
Mary Grush is Editor and Conference Program Director, Campus Technology.