What CIOs are learning with AI projects and AI outcomes

The gap between AI ambition and AI impact has never been more visible — or more costly. Across every industry, organizations have invested heavily in AI pilots, models, and platforms. Yet for many enterprises, those investments remain stubbornly confined to experimentation, never graduating into durable, measurable outcomes.

The question CIOs are grappling with today isn't whether to pursue AI — it's how to build programs that are trusted, secure, and built to scale. We asked some of the sharpest minds in digital transformation, enterprise strategy, and AI governance to share what they're seeing on the ground, and what they believe leaders need to relearn to make AI work where it matters most.

The answer starts long before the first model is deployed.

AI success starts with clarity

Helen Yu, CEO of Tigon Advisory Corp, makes the case that the biggest relearning for CIOs is getting the starting point right. "AI initiatives must start with the business objective and internal alignment," she says. "CIOs who anchor AI programs to measurable outcomes — cost reduction, customer experience, revenue growth — build natural guardrails that make trust, security, and governance easier to operationalize.

When the business problem is clear, the right data, the right controls, and the right oversight follow." The implication is direct: AI doesn't scale because you have great models. It scales because you have clear intent, clean data, and organizational accountability baked in from day one.

That foundation matters more than most organizations realize — because the failure mode, when it comes, rarely shows up where leaders expect it.

AI needs unified operating models

Sabine VanderLinden, Co-founder, Chief Executive Officer & Venture-Client Partner, Alchemy Crew Ventures, has watched this play out across regulated industries. "CIOs are relearning a hard truth: AI doesn't fail because the models aren’t powerful enough. It fails because organizations can’t operationalize their data. The pilots work. The scale doesn't."

The pattern she sees repeat itself is consistent: teams develop a clever use case, then discover the data foundation, governance rails, and human-agent workflows underneath were never designed for production. Orphan pilots live outside the P&L. Data readiness gets treated as an IT housekeeping task rather than a strategic asset.

And somewhere along the way, a quiet assumption takes hold — that AI is a feature you bolt on, rather than a fabric you redesign your business on. "Sustainable AI doesn't get built in the innovation lab," she concludes. "It gets built the moment the CIO and the COO stop running parallel agendas and start sharing one operating model."

This isn't a new problem. It's one the industry has encountered before — just in a different form.

AI demands operating model change

Dion Hinchcliffe, Vice President of CIO Practice, The Futurum Group, draws a direct line from past digital transformation efforts to where AI programs are failing today. "The biggest lesson from past transformation failures: technology without operating model change is usually just expensive theater. AI is no different."

His prescription for CIOs who want to break that pattern: treat data as a governed enterprise product, not a byproduct; redesign workflows around human + AI orchestration rather than copilots bolted onto legacy processes; and implement governance that measures outcomes, accountability, security, and cost in real time. "Most AI failures won't come from using AI," he warns. "They'll come from fragmented data, weak process intelligence, unclear ownership, and organizations still optimized for pre-AI ways of working."

And yet even the most well-designed operating model runs into a variable that no architecture can fully anticipate: people.

Scaling AI requires human adoption

George Westerman, Senior Lecturer, MIT Sloan School of Management, puts it plainly: "It's easy for AI innovators to forget that a model needs to live in a complex environment filled with people who may not want it." His first law of digital innovation holds that technology changes quickly, but organizations change much more slowly — and the real world is never as clean as a test environment.

"Innovating with AI may start as a technical problem," he says, "but scaling AI innovation is a deeply human one." His advice to CIOs: think about adoption from the very beginning, not as an afterthought. Who will need to embrace this innovation, and why would they — or wouldn't they? What can be done to bring them along at the start, so that scaling doesn't become a battle against the very people the technology was meant to serve?

What ties all these perspectives together is the recognition that AI outcomes aren't delivered by models alone — they're delivered by the infrastructure, governance, security posture, people, and operating models required to turn AI potential into business results.

Unified data and security power AI

Hem Nerkar, SVP & CIO, NetApp, offers an expert’s view of what that foundation needs to look like. On integrating AI-ready data solutions across hybrid and multi-cloud environments, he argues that the priority is treating data, security, and infrastructure as a unified platform — not separate layers.

"That means establishing a governed, single source of truth with real-time, trusted data, while embedding security, compliance, and auditability by design." This is operationalized through standardized platforms, AI-ready architectures, and reusable data services built for hybrid and multi-cloud deployment — with policy-driven access controls that scale globally without introducing friction to execution.

Security, in this context, can't be an afterthought either. Leading vendors, Nerkar notes, are shifting from reactive security to continuous, AI-enabled cyber resilience — integrating detection, protection, and response directly into the AI and data lifecycle. Zero-standing privilege, data classification, and behavioral analytics are becoming baseline expectations, not differentiators.

And when incidents occur, best-in-class vendors deliver integrated observability ecosystems with AI-driven anomaly detection, real-time monitoring, and automated response through SOAR platforms — tied directly to governance frameworks that ensure audit readiness, centralized logging, and closed-loop remediation. "The goal," he says, "is predictable, resilient operations with continuously improving risk visibility and response."

AI success through organizational change

Moving from AI projects to AI outcomes isn't a technology problem; it's an organizational one. As these perspectives make clear, the enterprises that will succeed aren't necessarily those with the most advanced models. They're the ones that anchor AI to clear business intent, redesign their operating models with human-AI collaboration in mind, govern their data as a strategic asset, and invest in infrastructure that can carry that work securely and at scale.

The shift from pilot to production is where AI programs prove their worth, and where the real relearning begins. Explore NetApp AI solutions to get your organization AI-ready.