The AI Multiplier Effect Starts With Your Data

Here’s a scene playing out in boardrooms right now: The CEO wants to know why the AI initiative is stalling. The CTO points to the model. The CISO points to the security perimeter. And the CDO quietly knows the real answer: the data is a disaster. Every new initiative kicks off with a six-to-twelve-month data-cleansing project before a single inference is run.

IBM’s 2025 CDO Study, drawing on 1,700 senior data and analytics leaders across 19 industries and 27 geographies, puts a number on the problem: only 26% of CDOs are confident their data capabilities can actually support new AI-enabled revenue streams. In a winner-takes-most market, that gap isn’t a technical debt footnote. It’s an existential risk.

The mandate has changed. The data hasn’t.

Two years ago, the CDO role was primarily about governance and compliance. AI changed that overnight.

The IBM CDO study documents this striking shift: CDOs now rank deploying data for competitive advantage as their top priority, ahead of governance and security. And 81% say their data strategy is now integrated with the technology roadmap,  up from 52% in 2023.

The execution is another story. 92% of CDOs say they must be business-outcomes-oriented to succeed. But only 29% strongly agree they have clear measures to determine the value of those outcomes. In other words, CDOs have accepted the mission but can’t prove they’re winning it.

The study also finds that CDOs who do outperform are 25% more likely to articulate how data priorities drive business results clearly. This suggests the articulation gap isn’t just a communication problem but a performance gap.

And it matters. 86% of CDOs say that when data leaders struggle to articulate their value, it jeopardizes the entire organization’s success. As Nestlé’s global head of analytics, data, and integration, Vikrant Bhan put it in the report: the CDO role is still evolving with no set templates. This is a significant challenge when balancing centralization and decentralization, as well as governance and delivery.

The silo is now an AI problem

The study’s most important finding isn’t buried in a chart. It’s the throughline across every section: fragmented data is the Achilles’ heel of enterprise AI.

When data lives in disconnected silos, AI agents — the autonomous software systems enterprises are betting their next decade on — either stall, hallucinate, or scale the wrong thing. Unlike human analysts who work around inconsistent datasets, AI perpetuates and amplifies whatever gaps it finds.

The CDOs outperforming their peers have figured something out: 81% of them now bring AI to data, rather than centralizing data for AI. The distinction may sound technical, but its implications are enormous. It means you avoid the cost and security exposure of relocating petabytes of sensitive data. AI workloads deploy wherever they’re needed, including the cloud, on-premises, and the edge. And data governance policies you spent years building don’t have to be rebuilt from scratch.

IBM’s own internal transformation offers a concrete reference point. The company’s CIO organization replaced seven disconnected data lakes with a unified hub, enabling a single query to pull from diverse systems that previously required laborious manual extraction. By Q1 2025, the platform had generated USD5.3 million in cost savings, offering a tangible proof point that architecture decisions have direct financial consequences.

From gatekeeper to growth driver

One CDO in the study reframed the role entirely. Wim Stolk, CDO of the Dutch Ministry of Economic Affairs, said: “I am not a Chief Data Officer at all. I am a Chief Trust Officer.”

It’s a statement worth thinking over. Because the study’s most striking cultural finding is that CDOs are consciously pivoting away from control, with more than two-thirds now saying their role is more focused on enabling use than preventing misuse. 82% go further, saying they’re actively wasting data if people can’t access it to make better decisions.

That’s a major shift in mandate for a function that spent its first decade building walls. 80% of CDOs say data democratization directly helps their organization move faster on AI. The question is how to open access without opening exposure. That’s precisely where the CDO-CISO relationship either pays off or falls apart.

The CDO–CISO alliance: The new power couple

The study calls the CDO-CISO relationship “data’s power couple.” This partnership can make or break an AI strategy. The framing is right.

A separate 2025 IBM IBV research of over 1,011 executives found that 87% say effective data security is essential for AI investments. But only half say they can actually protect sensitive data and prevent leakage across most AI use cases.

The gap between those two numbers is where agentic AI goes wrong. Autonomous agents don’t just query data; they traverse, combine, and in some cases exfiltrate it by accident. For CDOs deploying agentic pipelines, the question becomes about whether data lineage, jurisdictional controls, and access rights are machine-enforceable before the agents run.

This gap is especially acute for organizations operating across multiple regulatory regimes. 82% of CDOs say data sovereignty is now a critical aspect of risk management. That number will only increase as data residency laws proliferate and AI training pipelines cross borders.

The unstructured data blind spot

Here’s the number that should unsettle most CDOs: only 26% are confident their organizations can use unstructured data in a way that delivers business value. This matters because 78% of CDOs say leveraging proprietary data is a top strategic objective — and proprietary data is overwhelmingly unstructured. Call center transcripts, contracts, engineering reports, customer communications. The competitive moat most CDOs are trying to build sits in data they can’t yet use.

The challenge is sequential, not technical. Unstructured data must be structured, enriched, de-duplicated, cleansed, and stripped of sensitive information (including PII) before downstream AI can touch it. That’s not a one-time ETL job. It requires continuous, automated pipelines with lineage tracking baked in from the start.

MasOrange CDO Irene Yusta Martín explained in the study: technology has only recently given organizations the ability to treat unstructured data almost like structured data, including communications from call centers and messaging platforms. Most enterprises haven’t caught up to that capability, and the majority are sitting on a gold mine they can’t yet access.

Resilience: The most underestimated priority

Ask most CDOs which of the five focus areas (strategy, scale, resilience, innovation, growth) they’re most focused on, and resilience rarely tops the list. It should.

80% of CDOs say they’ve started developing diverse datasets to train AI agents. 79% say they’re still early in defining how to scale and govern them. That delta is where AI projects fail in production. An agent marketplace, where pre-vetted agents are registered, evaluated for security vulnerabilities, and deployed against standardized metadata, becomes the control structure that separates AI at scale from AI in chaos.

Centralized AI operating models are outperforming decentralized ones by 36% in ROI, according to the IBM CDO study. That’s a significant performance gap for an organizational design choice that too many enterprises are still treating as a cultural preference rather than a strategic decision.

The uncomfortable question

The study closes with a vision of the CDO evolving into an “AI systems architect.” CDOs will own enterprise-wide data reliability, interoperability, and AI-readiness as core functions and will be directly involved in revenue strategy and product development.

That’s a compelling trajectory. But the uncomfortable question the study asks is this: if only 26% of CDOs can currently support AI-enabled revenue streams, and 77% are struggling to fill key data roles, who exactly is going to build this architecture?

The talent problem is accelerating faster than the infrastructure problem. 82% of CDOs are hiring for generative AI roles that didn’t exist last year, but only 53% say their recruiting and retention efforts are actually delivering the skills needed. The CDO mandate has expanded dramatically. This is where a technology vendor like IBM, which underwent this transformation, can help — beyond platforms and solutions.

The AI multiplier effect is real. It multiplies value when the data foundation is solid. It also multiplies risk at scale and at speed when it isn’t.

Image credit: iStockphoto/Moor Studio