Every business leader reading the news right now sees the same headlines: record AI investment, sweeping automation promises, trillion-dollar forecasts. But a major new study released this week by NTT DATA tells a more complicated story — one that explains why so many expensive AI initiatives are quietly stalling.

The research, which surveyed nearly 5,000 senior decision-makers across more than a dozen industries, 30-plus markets, and five global regions, lands with a blunt headline: enterprise AI has hit a wall. And the wall isn’t made of bad models or poor strategy. It’s made of data.

The Numbers Tell the Story

The gap between awareness and action is striking.

More than 95% of respondents say private and sovereign AI are important to their business. But only 29% are prioritizing sovereign AI in any concrete, near-term way. That is not a knowledge problem. That is an execution problem — and it’s costing companies the competitive advantage they were promised.

Dig deeper into where the friction sits and the picture gets clearer. About 35% of Chief AI Officers say their top barrier is building, integrating, and managing AI models in private or sovereign environments. Nearly 60% of AI leaders cite cross-border data restrictions as a major challenge to deployment.

Put simply: the models are ready. The data pipelines aren’t.

Why This Is Happening

For most enterprises, the underlying data architecture was built for a different era. Centralized data warehouses, borderless data flows, and reporting-first infrastructure made sense when the goal was business intelligence dashboards. That infrastructure was never designed to support the locality, security controls, and jurisdictional constraints that serious AI deployment now demands.

Cross-border data restrictions are not a minor compliance hurdle. In the EU, the AI Act’s transparency rules take effect in August 2026. In regulated industries like banking, healthcare, and energy, data locality requirements are already baked into law. Running AI on data you can’t legally move — or store — requires a fundamentally different architecture than most organisations currently have.

NTT DATA’s research identifies a widening split in the market. Some organisations are redesigning their infrastructure from the ground up for control, locality, and security. Others are still trying to layer AI capabilities onto environments that were never built to support those requirements. The former are pulling ahead. The latter are generating impressive demos and stalled projects.

The Infrastructure Tax Nobody Talks About

There’s a phrase floating around enterprise data circles: the ETL tax. Every time you want to use data in a new context, you pay a toll in engineering time, pipeline complexity, and latency. For traditional analytics, that tax was manageable. For real-time AI agents making operational decisions, it’s often prohibitive.

The NTT DATA findings align with what practitioners are experiencing on the ground. You cannot build effective autonomous AI on a data foundation designed for quarterly reporting. The agents that are actually delivering ROI — the ones making real-time decisions in customer service, finance operations, supply chain, and risk management — are the ones sitting on top of governed, live, contextual data, not frozen snapshots in a warehouse.

What Leaders Are Getting Right

The research draws a clear line between organisations winning with AI and those still stuck in pilot mode. The winners share a few common traits.

They treat data infrastructure and AI governance as strategic requirements, not IT projects. They have invested in understanding exactly what data they hold, where it lives, what jurisdictions govern it, and how it flows between systems. They’ve built data catalogs, governance frameworks, and access controls before deploying agents — not after.

They also tend to have executive alignment on the difference between a proof of concept and production deployment. A demo running on a cleaned-up dataset in a sandbox environment tells you almost nothing about whether the same capability will work at scale, across time zones, across regulatory boundaries, with real messy data. The organisations learning this lesson early are avoiding painful and expensive rollbacks later.

What This Means for Your Business

If your organisation is in the 95% that understands sovereign AI matters but the 71% that hasn’t taken concrete action, the NTT DATA research should serve as a forcing function.

The regulatory environment is not getting simpler. The EU AI Act, national data residency requirements, and sector-specific rules are converging. Companies that wait for regulatory certainty before addressing their data architecture will find themselves scrambling when compliance deadlines arrive, rather than competing on the strength of their AI capabilities.

The good news is that the gap between current state and AI-ready state is bridgeable. It requires an honest audit of where your data actually lives, how it flows, and what governance controls you have in place. It requires prioritising data foundations alongside model selection — arguably ahead of model selection. And it requires treating AI infrastructure decisions with the same seriousness as ERP implementations or core banking system upgrades.

The businesses that get this right in 2026 will have a compounding advantage. The longer you run governed AI on live, contextual data, the smarter your systems get. The companies that delay are not just running slower — they’re falling behind on a curve that gets steeper over time.

---

For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.