If your enterprise AI “stack” has no single owner, you may not just be disorganized. You may be flying blind.

VentureBeat’s Pulse Research on the enterprise AI control gap finds a blunt mismatch between growth and control. Just under three-fifths (58%) of enterprises are net-adding AI initiatives over the past 12 months, with “expanding significantly” the largest single posture. But when it comes to visibility in production, confidence doesn’t translate into systems. Only 10% have active monitoring and alerting in place to detect an AI model drifting, behaving unsafely, or failing to complete tasks correctly. Even among the 40% who say they are very confident they would detect a failing model, the overwhelming majority of that confidence is based on manual human review (30%), not automation.

That’s the heart of the control gap: the ambition and spend machinery is running, while the “see it, own it, stop it” machinery is stuck in committee. And the reason is not that organizations lack AI. It is that they lack a governance center of gravity across a contested field of platforms.

In fact, the surface is contested in a way that makes governance hard by design. VentureBeat asked how many enterprise platforms currently claim to be the organization’s “primary” AI layer, with examples like ERP, EHR, ITSM, productivity suites, and data platforms each positioning themselves as the center of gravity. Almost no one gets a clean single answer. Adding the two multi-platform bands, 85% of enterprises have at least two platforms each claiming to be the “primary” AI layer, and more than a third (36%) describe an open four-way-or-more contest. Only 8% have consolidated to a single layer, and another 6% have not even mapped the question.

Think about the operational reality hiding behind those numbers. Each platform brings its own AI, its own controls, and its own assumptions. That means the rules for what “good” behavior looks like, where changes are detected, and how incidents are escalated can vary dramatically across the stack. Without agreement on a single governing center, cross-platform governance becomes less like a process and more like negotiation. You can’t enforce standards when you cannot even name who the standard is for.

VentureBeat’s findings point to ownership as the missing glue. When asked who is actually responsible for governing AI behavior across those platforms today, the headline answer sounds reassuring but the distribution undercuts it. Only a third (38%) say a central team governs AI today. A fifth (20%) say each platform team governs its own independently, and a fifth (21%) say ownership is unclear or contested between teams. Another 19% say no one has addressed it at all. Accountability fragments even further when respondents were asked which role holds primary accountability: CIO/CTO/CISO leads at 27%, a Chief AI Officer or equivalent at 22%, and 17% say no one holds formal accountability yet.

This is where board-level questions start to matter. If governance is an org-chart aspiration rather than an operating reality, then issues will show up late, and often in the worst place: in production, after users feel the pain. VentureBeat’s report describes exactly that detection gap. While 40% say they are very confident they would detect a model drifting, behaving unsafely, or failing in production, only 10% back that with active monitoring and alerting. More than a quarter combine the two reactive answers: 8% report no systematic visibility, and 19% would hear about problems first from end users.

The financial and operational stakes are already arriving through the cost channel. In the same survey, just under half (49%) name shadow AI, described as unauthorized agentic pipelines run on corporate cards outside central oversight, as their most severe control failure. Another 25% have been hit by a runaway “infinite loop” agent bill. Put differently: the control gap is not only about safety or model quality. It is about money moving without the enterprise’s blessing.

For context, VentureBeat ran this as part of its ongoing Pulse Research series, focused on governance, observability, and cost control across multiple AI platforms. Responses are filtered to organizations with 100 or more employees and, for this cut, exclude respondents selecting “Other” as their job function, leaving a base of identifiable roles (n=145). It is drawn from a single Q2 2026 wave (June), with sample tilt toward mid-market and lower-large bands: 100-499 and 500-2,499 employees (23% each). By role, it skews senior and technical: consultants and advisors (20%), CIO/CTO/CISO (18%), directors of engineering/IT (14%), product and program managers (13%), and enterprise architects (12%). The findings are directional, not a probability sample, but the patterns are hard to ignore.

The second-order implication for leaders is that regulatory pressure is likely to find you even if you do not feel ready. When systems cannot reliably detect drift, unsafe behavior, or task failures, accountability becomes reactive. And when spend is controlled by shadow pipelines and infinite loop incidents, internal controls start to look like they are auditioning for an audit. The strategic stakes for every exec in the room are the same: you can keep expanding AI portfolios, but the governance and observability layer cannot keep being the slower-moving partner. Otherwise, you end up managing AI incidents by hand, across a contested set of platforms, with costs that can quietly compound faster than visibility can keep up.