Is enterprise AI actually delivering, or is the industry’s favorite new statistic — the “data readiness gap” — mostly a sales hook for the vendors measuring it? The evidence in Dun & Bradstreet’s own numbers suggests both things are true at once. The company’s Q3 2026 AI Momentum Survey, covering 10,000 businesses in 32 countries and released July 28, 2026, finds 76% of firms reporting at least some AI ROI (48% “pockets,” 28% “broad or strong”) and 34% scaling AI into production. That’s real, incremental progress. But the same release pegs “fully ready” enterprise data at just 6%.
Worth noting: cio.com’s write-up of what appears to be an earlier wave of this same tracker put the readiness figure at 5% and ROI-reporting at 91% combined (67% “early signs,” 24% “broad”). Whether that’s a prior quarter or a different cut of the panel isn’t specified in either release, but the gap between waves — data readiness barely moving from 5% to 6% while ROI framing shifts by double digits – is a tough pill to swollow.
Outside of D&B – SAP and Oxford Economics’ Value of AI Report, also circulating this year, found only 3% of leaders “fully prepared” for agentic AI despite 83% seeing high potential — a strikingly similar structure from a different vendor with its own AI platform to sell.
“The challenge now is that AI adoption has outpaced data readiness. That is why only a few organizations have turned pilots into P&L-level ROI. Today’s frontier models are extremely capable, but getting the context right is the key to effectiveness. Grounding AI in verified information, so facts can be confirmed and integrations can be established, is fundamental to adoption of agentic workflows,” said Gary Kotovets, Chief Data and Analytics Officer at Dun & Bradstreet. “That is the opportunity that D&B is built to address. Our identity infrastructure grounds AI in real, verified business context, minimizing hallucinations and giving organizations the confidence to let their agents work for them. The D&B Commercial Graph™ pre-resolves crucial business context for AI, so you don’t have to burn tokens rediscovering and reconciling across departments.”
