Over the last year, I’ve had dozens of conversations with companies that are “doing AI.” They’ve launched pilots, experimented with models, and invested heavily in tools – an average of $400k this year. And yet, many of these organizations struggle to demonstrate real impact or articulate a clear ROI. In fact, by the end of 2025 at least 50% of generative AI projects were abandoned after proof of concept, in part due to poor data quality. 

I’ve seen this pattern before. 

When cloud adoption took off a decade ago, moving workloads wasn’t the hardest part. The real challenge was everything that followed: reliability, disaster recovery, cost control, operational efficiencies and organizational realities. That delta between early excitement and lasting value is where most technology initiatives struggle and it is what determines their success and impact. 

AI is no different.

In 2026, business leaders are occupied with everything from trying to predict the next outage to which AI model will win the compute wars, but those are less important than the durable moat you already own: your data. 

Companies struggle to implement AI effectively because systems are built on top of data foundations that were never designed to support them. The issue isn’t a lack of ambition or investment, it’s the underlying infrastructure. 

If cloud migration is like performing a heart transplant surgery on a patient while they are running a race, AI enablement is rewiring that runner’s nervous system into a supercomputer, mid-race. Making that leap requires companies to be able to access, trust, and actually use their data. AI can’t deliver value without clean, well-governed data flowing.

The Same Problem, Showing Up Again

When I founded CloudEndure in 2012, cloud migration was the urgent problem teams were trying to solve. I spent years working directly with customers and sitting down with infrastructure teams during migrations, outages, and recovery events. Data was technically protected, and that was enough at the time. 

Existing backup systems did their promised job in storing copies and aiding audits, but they locked data away and made it difficult and costly to access, reuse, or analyze. Every new initiative from analytics to disaster recovery started by rebuilding the same pipelines from scratch. Data was fragmented across accounts, regions, and providers, and every team paid a premium just to get access to information they already owned. 

Today, AI has exposed that problem again. Most AI efforts don’t fall short for a single reason, but again and again, data emerges as a key constraint. Even strong models can’t deliver value when the underlying data is scattered, incomplete, or locked in systems never designed for analytics or reuse. When data can’t move easily, AI stalls. The runner needs nutritious food.

Why We are at a Turning Point

In 2026, the need for accessible AI-ready data stops being a background infrastructure problem and becomes a center-stage, strategic one.

Data volumes are growing faster than teams can realistically manage with legacy approaches, and expectations for AI-driven insights are skyrocketing. At the same time, failures and outages – whether caused by cloud disruptions, security incidents, or operational errors – are more frequent and more expensive. With operational downtime costing an estimated $2M per hour, companies saw massive annual losses from IT outages 2025.

The way an organization stores and accesses its data determines what it can and can’t do. Teams that modernize their data foundations gain flexibility and can experiment with AI using data that’s already governed, compliant, and available.

On the flipside, teams that don’t modernize will stay stuck. They’ll run pilots that never reach production and duplicate efforts across departments. Critically, they open themselves up to costly disruptions and discover the data they need isn’t accessible when it matters most.

Looking Ahead: Data is Your Moat

I’m seeing a clear pattern: teams that get real results make a few key shifts. They treat data infrastructure as part of their operating system, align storage with business needs like speed and resilience rather than just archival safety, and confront the gaps in legacy systems where data exists but is effectively unusable.

I left AWS and co-founded Eon because this problem never really went away, it just evolved. I’ve learned that backup data becomes something companies store for emergencies instead of something they can use every day. 

Solving this inefficiency is now unavoidable. Not because AI is new, but because it finally forces a hard look at the foundations underneath it. Executives are recognizing that buried backups aren’t just lost opportunities – they’re wasted investments, until companies adopt systems that make them truly accessible.