Most businesses are investing in AI. Far fewer can prove it’s working.

That’s the central finding from Grant Thornton’s 2026 AI Impact Survey, which polled 950 business leaders across 10 industries between February and March of this year. The firm calls the result the “AI proof gap” — a widening disconnect between what companies say they’re doing with AI and what they can actually demonstrate.

The headline number: 78% of business executives say they lack strong confidence they could pass an independent AI governance audit within 90 days. This isn’t a question of whether they’ve deployed AI. Most have. The question is whether they could show it’s operating responsibly, consistently, and in alignment with business goals.

The Divide Is Already Showing in Revenue

The survey reveals a stark performance split between companies that have fully integrated AI and those still running pilots.

Organizations with AI fully integrated into operations are nearly four times more likely to report meaningful revenue growth than their counterparts — 58% vs. 15%. That gap isn’t hypothetical. It’s showing up in earnings now.

The difference isn’t just about having more AI tools. It’s about integration depth, data readiness, and whether teams are actually using AI as infrastructure rather than a bolt-on experiment.

The Data Readiness Problem

One of the clearest blockers: the underlying data isn’t ready for AI.

Fifty-five percent of CIOs and CTOs said fewer than half of their core applications are AI-ready. If the data that feeds your AI systems is fragmented, siloed, or poorly structured, no amount of investment in frontier models fixes that.

This is a pattern Enterprise DNA has seen directly through its work with hundreds of organizations. Teams that accelerated their data literacy in the 2020s are the ones now able to move fast with AI. Teams that skipped that foundation are stuck patching data pipelines while their competitors scale.

Who’s Being Left Behind

The survey also surfaces where AI adoption is stalling internally. Frontline employees were flagged by 37% of respondents as needing the most support to implement AI. Middle managers came in second at 30%.

This is the part that often gets missed in boardroom-level AI discussions. Executives set strategy. Vendors sell tools. But the gap between a pilot that worked in a controlled setting and actual business transformation lives in whether people in the middle of the organization understand what the AI is doing and can work with it effectively.

Only 40% of organizations surveyed said they were well-prepared to handle the privacy and security challenges AI creates. That’s a compliance exposure waiting to surface — particularly as the EU AI Act’s August 2026 deadline closes in.

What This Means for Business

The AI proof gap has a practical implication: companies that can’t demonstrate governance aren’t just at regulatory risk. They’re also less likely to see returns.

The Grant Thornton data makes the connection explicit. Organizations with AI integrated into operations — not just piloting it — are more than three times more likely to see revenue growth. The gap compounds over time as integrated organizations learn faster, deploy faster, and build institutional knowledge around what works.

Closing the gap requires three things most organizations are underinvesting in:

1. Data foundation. You can’t govern what you can’t measure. AI systems built on clean, well-structured data produce outcomes you can explain, audit, and improve. Organizations that have invested in data infrastructure are better positioned to move from pilots to production.

2. Workforce capability. AI governance isn’t just a policy document. It’s distributed judgment — thousands of daily decisions about when to use AI, how to validate outputs, and how to escalate edge cases. That requires people who understand AI at a working level, not just in theory.

3. Deployment accountability. Someone has to own outcomes. The organizations reporting strong AI ROI in this survey almost universally had clear ownership structures — people responsible not just for deploying AI, but for what it produces.

If you’re in the 78% that couldn’t confidently pass an AI governance audit today, the question isn’t whether to act. It’s where to start.

For teams looking to build the data and AI foundations that underpin confident, defensible AI deployment, Enterprise DNA’s learning platform offers structured upskilling across Power BI, Python, SQL, and AI — the practical skills that separate organizations who can prove their AI is working from those who are still hoping.

The practical next step is the free Working With Claude field guide. Thirty-two pages covering the ecosystem, Claude Code, and how to govern a rollout properly. Get your copy.