During a recent conversation on The Executive Outlook, Dr David Marco shared his thoughts on AI governance and how strong metadata management, data governance and data quality lay the foundation for responsible and scalable AI. With over 60 successful implementations and years of hands-on leadership experience in metadata and governance, his perspective is grounded in real-world enterprise projects, not just theory.
Dr Marco’s core belief is simple: AI cannot scale responsibly without trustworthy data foundations. Many organizations are excited about AI, but the real bottleneck is often messy, inconsistent, and poorly governed data. In this article, we explore how he connects metadata, KPIs, and use cases to build AI that is safe, compliant, and usable in the real world.
The Career Lesson That Changed Everything
Dr Marco’s journey into data governance didn’t begin with a planned career. Instead, it started with a decision to step forward early. While working as an ETL architect in the early days of data warehousing, he finished his work ahead of schedule and was asked to take ownership of metadata. At the time, there were very few resources on metadata, and the subject lacked much formal guidance. Rather than shy away from this unclear space, Dr Marco embraced it, researched what he could and began shaping a more structured way of thinking about metadata management.
This experience became a lifelong habit for him: raising his hand before the role was fully defined. He believes that this proactive mindset, which he applies in his work with AI governance today, is the difference between average results and meaningful transformation. Dr Marco sees it as a career-building lesson: successful careers are built on “empowerment over fear”, stepping up before you feel fully ready and learning by doing in the real world.
Prefer to listen on the go? Tune in to the full podcast episode on Spotify below:
The Real-World Impact of Bad Data on AI and Operations
One of the strongest aspects of Marco’s message is how he ties data governance to real-world consequences. He recalls the stress of system failures, late-night calls, and high-cost firefighting. These experiences gave him a deep respect for the damage bad data can cause and the operational chaos it creates across the enterprise.
This is why AI governance excites him. When companies start exploring AI at scale, the first concern they often hear is: “Our data isn’t ready.” To Marco, this is not a new problem. It’s the same foundational challenge that data governance and metadata management were created to address, now amplified by AI.
Watch the full conversation on YouTube by clicking the link below:
A Life-Critical Case Study: Data Governance for Cancer Analytics
Dr Marco shares a compelling example from a major hospital that wanted to build an enterprise data warehouse for cancer analytics. The stakes were unusually high because decisions powered by these analytics could directly affect patient outcomes.
Instead of immediately agreeing to build the data warehouse, Dr Marco advised the hospital that they were not ready. The foundational work needed to be done first. He emphasized three key pillars that must be in place before building high-stakes analytics or AI systems:
- Metadata management
- Data governance
- Data quality
He defines metadata as the technical backbone that explains the who, what, when, where, how, and why of enterprise data. Without it, organizations lack control, clarity, and consistency.
The hospital accepted this approach, and the result was not only a successful platform but also a measurable real-world impact, including reduced mortality rates. This case proves that data governance ROI is more than just cost savings; it can save lives.
How to Start Metadata Management for AI and Data Governance
When it comes to starting metadata management, he begins by focusing on business value, not just the tools. He emphasizes that metadata, governance and AI initiatives are major investments, often costing millions over time for large enterprises and government agencies. Without clear baseline KPIs, leaders risk inviting scepticism and unclear outcomes.
Only after clarifying what success looks like does Dr Marco recommend selecting tools that can support metadata management, governance, data quality and AI governance in a more integrated way. His advice is to avoid purchasing fragmented solutions and, instead, focus on a unified approach that reduces integration complexity and improves adoption.
A Real Enterprise Win: Finding Hidden ETL Waste
Dr David Marco shares an example from Micron, where metadata management helped uncover surprising inefficiencies in the ETL (Extract, Transform, Load) environment. After scanning the data landscape, they discovered that:
- 62% of ETL jobs were redundant or feeding sources no longer used
- Nearly 4,000 jobs could be turned off.
This not only reduced operational strain but also lowered costs, reinforcing Marco’s wider message: metadata gives organizations control over blind spots that quietly drain budgets and productivity.
The “Big Rocks” Way to Build Data Infrastructure
After establishing metadata management, Marco recommends expanding governance and quality through a focused, strategic approach. He advises organizations to prioritize critical data elements and the most mission-critical systems first, rather than trying to fix everything at once. Whether advising a mid-size manufacturer or a large federal agency, the approach remains the same: target systems that have the highest impact and build out from there.
The AI Adoption Chasm: Expectations, Fear, and Data Reality
Dr Marco describes the current AI landscape as a gap between unrealistic expectations and enterprise reality. Many executives assume that AI should be cheap and easy, given the consumer tools that seem simple to use. However, enterprise AI is fundamentally different; it’s more expensive, complex, and heavily reliant on clean data, clear governance and trained teams.
He acknowledges the fear organizations feel about AI accuracy and reliability, but he also believes the industry is maturing quickly. Frameworks like the NIST AI Risk Management frameworks are emerging to provide organizations with structure and confidence as they navigate the complexities of AI deployment.
Just as data warehousing once felt unfamiliar before becoming mainstream, Marco believes AI will follow a similar path. As standards strengthen and best practices become more widely understood, AI will become ubiquitous in organizations. From his conversations, many companies plan to focus on strategy and foundation building in the next few years, then scale their AI implementations once these basics are in place.
The KPIs That Actually Matter
When choosing KPIs for data and AI programs, he emphasizes the need to align them with business objectives and domain-specific needs. Using the hospital example, he highlights outcomes such as lowering mortality rates, reducing hospital stay durations, and improving treatment protocol accuracy.
Dr. Marco’s key insight is simple: data leaders should not define these KPIs in isolation. They need to consult with domain experts who live the problem daily and structure data programs around those real-world goals.
What to Avoid When Starting AI
Dr. Marco’s advice on early AI initiatives is straightforward. He recommends staying use-case-driven, keeping the scope manageable, and avoiding unnecessary complexity in the early stages. As he puts it, “Don’t try to boil the ocean.” He also warns that scanning every system in an organization doesn’t automatically improve AI accuracy.
Here are his practical guardrails:
- Start with high-value, realistic use cases.
- Avoid scanning low-quality side systems early.
- Prioritize systems of record first.
- Expand the scope only after proving value.
A Simple Framework to Prioritize High-Value AI and Data Use Cases
Dr Marco’s framework for defining AI and data use cases is based on a logic that leaders can apply in any industry. He suggests evaluating opportunities in this order:
- Business objectives or mission
- Real pain points that create risk, cost, or decision uncertainty
- Revenue growth opportunities
- Cost reduction and duplication cleanup
- Creative ideas, as long as they tie back to value
He supports this with real enterprise stories, such as the financial services case where a data steward caught a potential error just before regulatory reporting—avoiding a public violation and massive reputational damage. In another example, sensitive PII was found duplicated across uncontrolled environments, highlighting the urgent need for data governance.
Each example reinforces the same truth: governance is not bureaucracy. It’s what makes AI safe, scalable, and trusted.
Final Thoughts
Dr David Marco strikes a rare balance of ambition and discipline. He accepts AI innovation; he insists that AI must be built on strong foundations. His perspective is rooted in real-world experience, where technology succeeds only when people trust the data behind it.
His message is clear: start with metadata, anchor programs in measurable KPIs, focus on the big rocks, and scale AI only when the foundation is strong enough to support it.
For more inspiring stories of leaders shaping the future of data, AI, and strategy, stay tuned with The Executive Outlook.
Editor Bio
I’m Isha Taneja, serving as the Editor-in-Chief at "The Executive Outlook." Here, I interview industry leaders to share their personal opinions and provide valuable insights to the industry. Additionally, I am the CEO of Complere Infosystem, where I work with data to help businesses make smart decisions. Based in India, I leverage the latest technology to transform complex data into simple and actionable insights, ensuring companies utilize their data effectively.
In my free time, I enjoy writing blog posts to share my knowledge, aiming to make complex topics easy to understand for everyone.
