CDAOs tackle roles, data, and agent coordination

Organizations reevaluate operations, leadership roles, and governance frameworks for data and AI.

The rapid acceleration of AI is forcing a re-evaluation of organizational design and enterprise leadership roles. Across organizations, there is a significant tension between centralized (top-down) control and decentralized (bottom-up) innovation.

Most leaders are finding that a purely top-down or bottom-up approach is insufficient, leading them toward a hybrid model. The bottom-up approach allows for rapid experimentation and empowers business users, but it often leads to duplicated efforts, rising costs, and a failure to scale complex solutions.

That said, operations are becoming more federated. Value creation, rapid prototyping, and localized AI innovation are increasingly embedded directly within the business units. This provides centralized guardrails while empowering the business to move at the speed required to capture competitive market advantages.

Role-wise, chief data officers are playing both offense and defense. The offensive role is business focused, handling use cases, adoption, value realization, and enterprise execution, while the defensive role is responsible for data foundations, data strategy, and quality of lineage.

This delineation in roles and responsibilities is necessary. Organizations are moving away from simply proving that they can implement AI, shifting their focus toward measuring the tangible business value generated by these technologies.

Centralized teams supply the underlying infrastructure and technical standards, but individual business units now hold primary accountability for investments and value realization. However, the decentralized rush to innovate can create internal friction. The intense pressure to deploy AI faster can lead to duplicated efforts or a chaotic environment where established operational lanes disappear, occasionally distracting from core, day-to-day business operations.

To address this, there is a sentiment that any successful AI adoption will also demand a cultural shift. Several CDAOs emphasiz the need for employee upskilling and compelling people to “fall in love with AI” to move from a state of fear to value creation.

AI’s growing influence on organizations also reverberates in the role of chief data officers. It is shifting, notable by a trend of aligning directly under the Chief Financial Officer (CFO). This reflects the growing imperative to tie data investments directly to financial performance and strategic value. Simultaneously, the rise of the Chief AI Officer introduces new complexities, as reporting lines for this emerging role remain highly variable and undefined.

"“Bottom up has a lot of gravity for true AI enablement and adoption.” Danielle Beringer, Principal, Advisory, Tech & Data

Data challenges spark a growing focus on knowledge engineering

The ultimate ceiling for any AI initiative is not the algorithmic itself, but the quality of the data feeding it. As organizations transition from AI pilots to scaled enterprise deployments, a stark reality is setting in: even the most advanced AI models cannot overcome the limitations of a poor data foundation. Consequently, data quality is experiencing a renaissance, driven by the impact that data reliability has on the accuracy and trustworthiness of AI outputs.

As many CDAOs put it, “Your AI is only as good as the underlying data.” 

Historically, enterprise data governance favored structured data residing in traditional databases. However, Generative AI thrives on unstructured data—documents, emails, and internal communications—exposing previous blind spots in governance frameworks that are now necessary to address. Case in point: perhaps the most valuable enterprise data is not written down at all. It exists as tribal knowledge locked within the minds of experienced employees. To truly contextualize AI and maximize its value, organizations must bridge this critical gap. 

The diversity of data has sparked a growing focus on knowledge engineering, an emerging discipline dedicated to capturing and codifying mostly unstructured data like human expertise so AI systems can leverage it. Establishing the right level of investment in knowledge engineering is now recognized as a vital step in making data AI-ready.

Despite the renewed focus on data quality, securing enterprise-wide accountability remains a stubborn challenge. For far too long, IT departments have served as the de facto proxy for data quality remediation. The consensus is that data quality is fundamentally not an IT issue; it is a business process and business ownership issue. True remediation requires a shift away from IT-led cleanup efforts toward full transparency and business-driven accountability.

To scale AI use cases, organizations are establishing robust, data-level governance frameworks that mandate strong partnerships between technical teams and the business. By packaging data into clear, business-owned data products, enterprises can enforce accountability directly at the source. Ultimately, realizing the transformative promise of AI requires organizations to stop viewing data quality as a backend technical chore and start treating it as a strategic, business-owned imperative essential for long-term competitive survival.

“With AI-ready data, the most valuable data may not be written. It’s in people’s heads. That knowledge will help contextualize data for AI.” Matteo Colombo, Principal, Advisory, Tech & Data Integration

The Shift from Static Dashboards to Dynamic, Agentic AI

The era of static reporting driving enterprise decisions is rapidly ending. While rumors about the death of the dashboard may be exaggerated, there is a seismic shift toward dynamic, agentic AI. Dashboards retain their utility for high-level, at-a-glance operational monitoring, but complex, conversational analysis is moving toward AI agents capable of contextualizing data and answering deep, ad-hoc business inquiries on demand. As proof of how quickly this shift is accelerating, some CDAOs are planning to cancel their business intelligence contracts within a year.

This shift reduces the time required to build AI solutions. Business units are increasingly empowered to rapidly convert their operational ideas into functional AI tools. 

“We built an agentic platform that allows for a community approach to agent development,” according to a CDAO with a CPG company. “We have around 100 agents deployed generating revenue or taking cost out in different areas.” 

However, this democratization of development can lead to a fragmented, multi-platform reality. Different departments often gravitate toward different foundational models, creating a sprawling ecosystem of disconnected applications.

Managing this multi-platform environment is an operational hurdle. Building solutions quickly, deploying, managing, and governing them across a unified enterprise architecture can be incredibly difficult, yet worthwhile. 

As an automative supplier CDAO articulated, “In addition to the monthly report, we have an ad hoc interface for asking questions. It’s completely addictive.” 

To stitch together cross-platform, cross-agentic governance, many organizations are exploring the concept of a centralized AI harness or a registry of registries. This shared infrastructure is deemed essential to manage agent skills, provide consistent enterprise context, and enforce security guardrails across disparate systems without stifling localized innovation.

Consequently, the agentic future is forcing a rethink in how enterprise data is packaged and delivered. There is a growing strategic distinction between traditional data products engineered for human knowledge workers and data products designed to flow directly into the AI execution plane. Data leaders are actively pivoting their focus to accommodate how data products are developed. 

The goal is to prioritize scalable, agentic solutions that assist and augment workflows, transitioning the enterprise from a posture of simply reviewing historical metrics to interacting with an intelligent, dynamic operational ecosystem. The reality is that prioritization and intent are easier said than done.

“I have 20,000 dashboards, but my competition isn’t another dashboard. It’s PowerPoint and Excel slowing implementation of a more scalable solution.” CDAO for an energy company.