Chloe Maraina is a powerhouse in the world of data science, known for her uncanny ability to transform massive, cold datasets into vivid narratives that drive corporate strategy. As an expert in Business Intelligence and AI governance, she has spent years navigating the murky waters of data integration and the ethical implementation of machine learning. Today, we sit down with Chloe to discuss the intricate dance of operationalizing AI governance within global enterprises. We explore the paradox of public skepticism versus corporate adoption, the strategic necessity of building influence over mere authority, and the surprising reality that a tiny, elite team can govern a multi-billion dollar AI landscape. Our conversation delves into how trust is built through technical enablement, the specific makeup of a high-impact governance squad, and why “win-win” partnerships are the secret weapon for any leader looking to move from theory to practice in the age of generative AI.
Public skepticism toward AI remains high, with many individuals reporting low levels of acceptance. How do you reconcile this widespread cultural hesitation with the fact that nearly ninety percent of organizations are already rushing to implement governance frameworks?
It is a fascinating and somewhat jarring disconnect that we are seeing in the market right now. On one hand, you have the general public where sixty-one percent of people are genuinely wary of trusting AI, and a staggering sixty-seven percent report only low to moderate acceptance of these technologies. You can almost feel that collective intake of breath whenever a new model is announced; there is a visceral fear of bias and a loss of agency. On the other hand, the corporate world is moving at breakneck speed, with about ninety percent of organizations already using AI actively working on governance. These companies realize that they cannot reap any real business value if the systems they build are fundamentally mistrusted by the people they serve. The reconciliation happens when we stop viewing governance as a series of “no” boxes to check and start seeing it as a way to instill that missing trust. By being transparent about how models perform and ensuring record-keeping is not just sporadic but rigorous, enterprises can bridge that gap between public fear and technological progress. It is about proving that responsible AI is not just a white paper concept but a practical, operational reality that can be managed even by a lean, dedicated team.
When a company’s AI portfolio is growing as rapidly as we have seen recently—doubling year over year in some cases—how does a small team maintain oversight without becoming a bottleneck for innovation?
Scaling oversight in an environment where the number of AI systems doubled every year until 2024, and even then grew by another sixty percent, requires a shift from manual checking to strategic enablement. If you try to put a single person in front of every new project, you will experience a total system stall, which creates a massive backlog of legacy products and heightens operational risk. The key is to realize that you do not need an army; you need a highly skilled, five-person strike team that focuses on creating consistency across diverse environments. When you have different teams using different tools and contexts, the documentation often becomes non-existent, leaving you to scramble to close gaps for regulated customers. We have found that providing proactive risk guidance and access to self-service tools allows the developers to do the heavy lifting themselves. By shifting the “control verification” to a process where the team provides evidence of mitigation, the governance team moves from being a police force to being a group of strategic advisors who ensure that the volume and complexity of the work do not outpace the safety measures.
You have mentioned that building influence is often more effective than exercising raw authority when it comes to governance. How do you practically create those “win-win” scenarios with developers who might see governance as a hurdle?
Building influence is all about finding the pain points of your peers and solving them in a way that happens to satisfy your governance requirements at the same time. I often look back at the early months of 2022 as a turning point for this strategy, where the goal became making compliance the path of least resistance. For instance, if you partner with a data science team to create a model documentation template, you aren’t just giving them more paperwork; you are giving them a tool they can take to customers pre-contract to prove the model’s efficacy. This upfront information sharing creates immediate trust with the client and helps close deals faster, which is a massive win for the business side. When governance delivers something of tangible value, like a streamlined API for bias testing, it creates a reputation for being a partner in success rather than a hurdle to be jumped. You want to be the team that helps people succeed in ways that align with the company’s goals, and once you have a track record of delivering those win-wins, your influence grows organically without you ever having to demand it.
If you were to build a “dream team” for AI governance with only five specialists, what specific roles and behaviors would you prioritize to ensure they can cover a global enterprise?
The composition of a small governance team is everything; you need a specific blend of technical grit and communication prowess. First, you need an architect who can take the high-level framework elements and actually wire them into the designs and processes used by other teams across the company. Then, you absolutely must have a model risk expert who deeply understands the regulatory demands, especially if you are dealing with major banks or regulated entities that expect detailed performance data. You also need a specialist dedicated to building and iterating on the technology solutions themselves—someone who can manage the technical debt and run those crucial development partnerships. To round it out, a communications and networking expert is vital to drive the messaging of “responsible AI” and ensure that the team’s thought leadership is actually reaching the right ears. Finally, the lead should ideally have a background in R&D and data science to maintain credibility with the technical staff. Beyond their titles, I look for two non-negotiable behaviors: curiosity and self-awareness. You need people who are transparent about what they don’t know and are hungry to figure out how these complex systems really tick under the hood.
Operationalizing ethics can often feel abstract, but you’ve advocated for very concrete tools like scorecards and bias-testing APIs. How do these tools change the day-to-day workflow for a product owner?
These tools take the “mystery” out of compliance and turn it into a structured, manageable workflow. It starts with a scorecard that a product owner completes before a single line of code is written or a contract is signed. This forces them to ferret out the hard truths: Do they actually understand the data they are using? Does the system have the agency to make autonomous decisions? If the scorecard flags a risk, we treat it as a real threat until proven otherwise, which shifts the burden of proof to a place of safety. To make this easier, we provide the tools—like a specific API for bias testing or mitigation requirements for large language model evaluation—so the owner isn’t left guessing how to fix a problem. It turns a vague requirement like “ensure the model is fair” into a series of technical actions that can be measured and verified. When you align the testing requirements directly with the tools you are distributing, you remove the friction that usually leads to sporadic record-keeping, ensuring that every project has a clear, documented trail of accountability.
What is your forecast for AI governance?
I believe that by 2026, we will see a massive shift where AI governance moves from being a specialized “side-car” function to being the foundational lynchpin of the entire enterprise software lifecycle. We are already seeing ninety percent of organizations starting this journey, but the next two years will be defined by the professionalization of the “AI Governance” role as a distinct, highly-coveted career path. As the complexity of large language models continues to grow, companies that rely on manual, ad-hoc reviews will simply fail to keep up with the market. My forecast is that “Enablement-as-Governance” will become the standard, where the most successful companies won’t have the largest legal teams, but rather the most integrated technical tools that make bias testing and risk mitigation as automated as a standard security scan. This work is not just urgent; it is the only way we will move past the sixty-one percent skepticism rate and actually start delivering the profound business value that AI promises in a way that feels safe and human-centric.
