CDAOs: The new mandate for AI value

Voice of the CDAO | Insight Series

As executive scrutiny increases, organizations are placing greater emphasis on demonstrating measurable AI impact. Chief Data & Analytics Officers (CDAOs) are now under a new mandate: deliver verifiable, CFO-certifiable impact. The conversation has pivoted sharply to proving performance, forcing leaders to tackle the deep operational friction that separates AI hype from bottom-line reality. Data leaders are rewriting the value narrative to prioritize competitive advantage. They are simultaneously addressing growing challenges around cost, governance, and risk through rigorous FinOps and automated governance. Most critically, they are architecting a new data-centric talent operating model, recognizing that reinventing the workforce is the ultimate key to unlocking the agentic future. The journey is complex, but the direction is clear: success is now measured in outcomes, not activity.

From ROI theater to competitive reality

The pivot requires a new mindset

CDAOs view high-quality, centralized data as the perquisite for moving beyond efficiency plays. Additionally, traditional ROI calculations often fail to capture the strategic impact of AI.

Across businesses and industries, it’s referred to as “ROI theater,” in describing how teams are forced to justify projects with speculative, hard-to-measure financial projections. The consensus is to pivot the conversation toward enabling competitiveness.

A powerful example of this is deploying AI in reviewing complex financial documents. It’s a task previously limited by a small number of specialized employees who could perform it. By automating the review process, the firm unlocked a significant new revenue stream. The lesson is clear: the most compelling value stories are often found not in broad efficiency gains, but in solving business problems that contribute to the bottom line.

This shift in thinking requires a new mindset. It’s often about tolerating some ambiguity in initial business cases for high-profile big bests. There is an age-old concern around analysis leading to paralysis.

The optics in the financial services sector typify discussions around competitive reality in the age of AI. “Instead of spending time trying to litigate a dollars-and-cents use case, it’s about answering, ‘Can we do something we couldn’t do before that’s going to make us more competitive?’” The result was a singular database for finance that served as an enabler for everything else. Without it, high-velocity AI development would be nearly impossible.

Treating AI less like a strategic capability is akin to a free-market system where managers make gut calls for their department, with the understanding that not all bets pay off immediately. This is particularly true for growth initiatives like churn prediction, where the value may trail beyond the normal expectations for ROI.

A data executive in the investment management industry shared the art of thinking bigger.

“We have gone through a grassroots effort to cultivate use cases that are game changers. Let’s forget about pilots. Game changers are the things we’ve wanted to do for 10 years. We just never had the capability. Now we do."

"Our initial value framework was around risk readiness and ROI. Now, we’re building an AI factory that can drive the organization forward.”

— Danielle Beringer, Principal, KPMG

Data governance delivers AI risk management

Frameworks track token usage down to the individual user

The conversation among data leaders is data governance is the central pillar of AI risk management. It’s about building technical and procedural guardrails to control how data is used by agentic systems in real-time.

How that manifests itself varies from company to company and data officer to data officer. A chief data officer in the technology arena shared their approach.

“We put in place an AI gateway that tracks access to LLMs internally or externally. We also have an MCP gateway. We have controls in place to leverage these gateways. We’ve also integrated FinOps through this whole process, so we’re able to track token usage down to the individual.”

The days of centrally funded, "black box" AI budgets are numbered. The rising tide of consumption-based pricing from vendors is forcing a shift toward granular financial operations (FinOps) and direct accountability within business units.

With that in mind, organizations are changing how they're doing their internal billing and overall ROI calculations. Firms are building frameworks to track token usage down to the individual user and application, tying the cost back to the original business case. This allows finance and business leaders to make informed economic decisions, turning a once unknown technology cost into a manageable operational expense.

This financial discipline is mirrored by an urgent focus on risk governance. Many organizations have established AI governance programs with distinct lanes for rapid experimentation and production-scale use cases. These programs integrate reviews across multiple risk domains—cybersecurity, privacy, compliance, and HR—and are increasingly automated.

Across the technology sector, here’s the favored approach: "We have controls in place, including relying on human-in-the-loop with clear guidelines on when to escalate decisions, preventing scenarios like an AI agent unilaterally issuing high-value customer vouchers.

The goal for CDAOs is ushering in real-time risk management. This is critical for deploying autonomous agents where human oversight shifts from being "in the loop" to "on the loop." To build trust in AI systems, firms are partnering with third-party AI underwriting companies to run tests, simulating how agents behave under duress. This rigorous validation is becoming the new standard for ensuring that as AI-driven processes scale, guardrails scale with them.

"Token counting is bean counting. It's not going to give us anything useful, but it will definitely tell us what is happening from a consumption perspective.”

— Swami Chandrasekharan, Head of AI and Data Labs, KPMG

Data-centric talent as the new operating model

Reinventing work and cultivating a hybrid workforce

The value of employees is increasingly defined by their ability to interact with and improve data. It represents a fundamental shift in talent strategy, where proximity to data and the skill to validate it are becoming more critical than where employees are located. Employees must understand the downstream implications of creating good or bad data.

This calls for fundamentally reinventing work and cultivating a hybrid workforce that can operate in an AI-augmented environment. In a nutshell, it’s a new operating model.

Every sector sees this challenge. “The problem is there are not enough people who understand business and the art of the possible with AI.”

This skills gap is a major impediment to moving beyond incremental improvements and achieving step-change transformation. To bridge this gap, organizations are adopting a dual approach: a top-down mandate for big bet projects paired with broad, bottom-up enablement.

This involves everything from running design-thinking workshops to creating AI-native workflows to developing an AI augmentation slide rule that maps existing roles to future-state capabilities. It ranges from simple augmentation to fully autonomous agents.

This role-by-role analysis helps prioritize which functions are ripe for AI-led transformation versus those better suited for traditional labor managed through global capability centers. Transformation isn’t about smooth sailing. There is a lively debate between "transition first" (offshoring for cost) and "transform first" (automating for efficiency). The consensus is that attempting both simultaneously is a recipe for failure.

"The pressure, in my personal opinion is that management is losing patience,” remarked a data and analytics executive for a consumer goods company. “If you're spending 11% of your budget on AI, management wants to know the impact. If it's not moving the needle, we risk losing support and losing momentum."

Leaders emphasize that success requires getting business VPs to take ownership of their data domains. When people’s primary roles shift from executing tasks to organizing data and validating AI-driven output, the entire operating model changes. It should foster a culture where creating high-quality data is not a bureaucratic chore but a critical enabler of the company's agentic future.

"Operational intelligence is going to be a competitive advantage. That is contingent on organizations knowing when to escalate to people. That's how you build your moat."

— Analytics executive, Consumer goods sector

Defining AI value beyond efficiency
As AI adoption matures, organizations are increasingly evaluating value through competitive advantage, growth opportunities, and new business capabilities rather than efficiency gains alone.
Balancing scale, governance, and accountability
The expansion of AI is placing greater emphasis on cost transparency, risk management, and governance while maintaining the flexibility needed to support innovation.
Preparing for a data-centric operating model
The growing role of AI is reshaping how work is organized, elevating the importance of data stewardship, workforce readiness, and the balance between human and machine-led activities.