--- title: "AI Capability Went Up This Year. Enterprise ROI Didn't Move At All." date: 2026-08-06T09:00:00Z author: Deepankar Bhadrasen url: https://truehorizon.ai/news/domino-ai-roi-capability-gap-enterprise-report description: Domino Data Lab surveyed 639 enterprise AI leaders and found capability rising while ROI stays exactly where it was a year ago. --- # AI Capability Went Up This Year. Enterprise ROI Didn't Move At All. > Domino Data Lab surveyed 639 enterprise AI leaders and found capability rising while ROI stays exactly where it was a year ago. On July 21, 2026, Domino Data Lab released its [Fifth Annual Enterprise AI Report](https://www.prnewswire.com/news-releases/ai-roi-fails-to-outpace-spend-for-57-of-enterprises-unchanged-since-2025-even-as-93-now-report-improved-production-302830222.html), and the headline finding is a genuine contradiction. The survey, fielded independently by BARC Research across 639 senior AI leaders at companies with $100 million or more in revenue, found that 93% report improved AI production capability in 2026, up from 88% in 2025. Whatever enterprises are building, it is measurably working better than it did a year ago. The number that should worry a CFO sits right next to it. 57% of those same enterprises report ROI that fails to outpace their investment. That figure hasn't moved. It is identical to 2025. Two years of capability gains, and the share of companies whose AI spend is actually paying off is exactly where it started. ## The milestone that stopped mattering "Getting a model into production used to be the milestone that mattered," said Thomas Robinson, COO at Domino Data Lab. "Our research shows that's not enough anymore." That is the report's core claim in one sentence. Capability was never the bottleneck people assumed it was. The ROI plateau held steady for two straight years while capability climbed five points, which rules out "the model isn't good enough yet" as the explanation. ## The regional split nobody's talking about The report [breaks the 57% figure down by region](https://www.carriermanagement.com/news/2026/07/22/290293.htm), and the split is sharp. In North America, 51.1% of respondents report ROI stuck at or below investment level. In the UK it's 66.9%. In continental Europe it's 67.0%. Three regions running the same generation of AI tools, with return outcomes that differ by 16 points. A universal maturity story would not produce numbers like that. Something regional, whether procurement process, talent availability, or governance posture, is doing real work here. ## The governance number that explains the other two The most specific finding in the report may also be the most useful one. Only 43% of organizations running agentic AI have it in governed production. The other 41% are either piloting or actively scaling agentic AI with no governance in place at all. Shawn Rogers, CEO of BARC US, put the fix plainly: "Govern early, and build the applications that turn AI into something business users can actually use." Read against the ROI numbers, that is not a side note. Ungoverned agents are exactly the kind of deployment that produces capability without measurable return. They run, they do work, and nobody can cleanly attribute a dollar figure to what they are doing. ## The last-mile gap Domino names a second, related problem. Only 34% of organizations report consistent AI access methods across business units, the single most common answer being a patchwork that varies department to department. A model that works in production but that business users cannot reliably reach is a model that cannot show up in anyone's ROI calculation. Capability sits in the lab. Value sits wherever a business user could actually get to it, and for most enterprises, that access is inconsistent by the report's own numbers. ## How True Horizon does it This is the exact gap we build for. We do not treat a working model as the finish line. We build the governance layer that lets an agent scale without becoming ungoverned, and the access layer that gets a working model in front of the business users who are supposed to be creating value with it. Capability without access is a demo. Access without governance is a liability. The report is describing enterprises that solved one and not the other. ## What to do now Pull your own numbers apart the way this report does. Separate "is the model better than last year" from "did ROI move," because Domino's data says those two questions now have completely different answers. Audit every agentic AI deployment you have and sort it into governed production versus ungoverned pilot or scale, since that split alone predicts a meaningful share of the ROI gap. And check whether access to your working AI tools is actually consistent across business units, or whether, like most of the companies in this survey, it varies by team in ways nobody has mapped. If you want to know whether your own AI capability is quietly outrunning your ROI, [take our AI assessment](https://truehorizon.ai/assessment) and we'll show you where the gap actually sits. Source: https://truehorizon.ai/news/domino-ai-roi-capability-gap-enterprise-report