The three thinks that AI leaders get right and the 3 that they get wrong. (Photo by Kirill KUDRYAVTSEV / AFP) (Photo by KIRILL KUDRYAVTSEV/AFP via Getty Images)AFP via Getty ImagesEnterprise AI has cleared the adoption hurdle and stalled on the proof hurdle, according to Plug and Play’s 2026 Enterprise AI Strategy Pulse Survey, released today. Seventy-four percent of the world’s largest enterprises now run at least one AI solution in production and 93% are piloting or further along. Half of those production-stage companies cannot tell you whether any of it worked. The sample skews to Fortune 500 and Forbes Global 2000 companies, so this is the top of the market, not the long tail."Enterprise AI has crossed the deployment threshold, but not the value threshold. Nearly three-quarters of the enterprises we surveyed have AI in production, yet half still cannot consistently measure ROI. The next chapter is less about the models themselves and more about the operating foundations around them: better data, governance, integration, ownership, and measurement. Enterprises that pull ahead will be those that consistently translate production deployments into tangible business outcomes," Amit Patel, Ventures Partner at Plug and Play, told me. Here’s the overall Enterprise AI Gap from all the reports that are in the market now. Multiple Reports including Plug and Play.Sandy CarterHere is what the top cohort is getting right, and where it is going wrong.Right: Enterprise AI Leaders Stopped Debating And ShippedTwo years of pilot purgatory has ended. Only 5% of these organizations call themselves AI native, but 37% run AI inside a single function and 32% operate it across several. Deployment is no longer the constraint, and that is a real accomplishment that gets lost in the ROI hand-wringing.MORE FOR YOURight: Enterprise AI Buyers Choose Defensibility Over DemosNinety-two percent of respondents rank data privacy, explainability and compliance as their top vendor selection factors, ahead of performance at 74% and flexibility at 53%. That ordering would have been unthinkable in 2024, when benchmark scores sold deals. Enterprise buyers have learned that a model they cannot explain is a model they cannot defend to a regulator, a board or a customer.Amit Patel, Partner at Plug and Play, shares valuable insights from the latest 2026 study.Plug and PlayRight: Enterprise AI Ownership Moved Into The BusinessOnly 21% of AI ownership now sits with a CAIO, AI lead, or center of excellence. Thirty-seven percent sits with functional or line-of-business heads, and BCG found 72% of CEOs are now the main AI decision-maker. This is correct in principle. Value is created inside the function, not inside IT, and the person accountable for the outcome should control the tool.Wrong: Enterprise AI Ships Without A BaselineThe narrowest deployments are the worst measured. Among companies running AI in a single function business function, the earliest and simplest stage of production, 74% three-quarters say ROI is too early to measure or is not tracked at all. That is worse than the topline half, and it is backwards from what you would expect, since a single function should make attribution easiest. A use case that goes live without a pre-deployment metric is unmeasurable forever, because there is nothing left to compare against.Larger datasets confirm this is not a small-sample artifact. KPMG’s June 2026 pulse of more than 2,145 leaders across 20 countries found only 7% report established AI ROI. Deloitte's study of 3,235 leaders found 66% report productivity gains but only 20% see AI-driven revenue growth.Wrong: Enterprise AI Authority Moved Without The AccountingThis is the mirror image of the third item AI leaders got right. Central IT spent decades learning to attribute spend to outcomes. A line-of-business head has a P&L but rarely a shared measurement framework, and almost never one that lets a CFO compare a marketing deployment against a supply chain deployment. Authority moved out of the technology function faster than the accounting discipline followed it, and the measurement gap is the bill for that.Wrong: Enterprise AI Is Measured On The Wrong CalendarIDC and Microsoft measure an average return of $3.70 per dollar spent on generative AI, with a median time to positive ROI of 14 months. Most enterprises review on four quarters. That mismatch kills projects on their way to a positive return and rewards the ones that produce a fast, shallow win.The same short term viewshows up in data spending. Data foundations remain the top blocker to scale at 71%, and they stay elevated at every stage of adoption rather than fading as companies mature. Most enterprises still fund data work as project overhead instead of a standing line item. Integrate.io found strong data integration produces 10.3 times ROI against 3.7 times for poor connectivity.The One Enterprise AI Move For The Second Half Of 2026Set up an AI metrics standard, owned by the CFO. Without it, every line-of-business leader will create a defensible framework for their own function, but none of those frameworks will be comparable or roll up across the enterprise. That is how companies end up with ownership sitting in 37% of the business and no unified view of value. Proof is what will carry Enterprise AI investment through the 2027 budget cycle.