OutSystems Surveyed Nearly 1,900 IT Leaders. 96% Are Already Running AI Agents. 94% Worry They Can't Control What They Built.

October 28, 2026 · Josh Agusti, Founder, Aiven Consulting Group

Ask enterprise IT leaders whether their organization is using AI agents, and the answer is close to unanimous. OutSystems' 2026 State of AI Development report surveyed nearly 1,900 global IT leaders between December 2025 and January 2026, and 96 percent said their organization is already using AI agents in some capacity. Ninety-seven percent are exploring agentic AI strategies that span the whole enterprise, not just a single team's pilot.

What's harder to find is confidence that anyone has a handle on what's actually been built. Ninety-four percent of the same IT leaders said agent sprawl, the proliferation of agents built by different teams for different purposes with no shared oversight, is increasing complexity, technical debt, and security risk inside their own organization.

Why adoption outran the ability to manage it

The pattern tracks with how agentic AI actually spread inside most enterprises. A sales team stood up an agent to draft proposals, a finance team built one to flag invoice exceptions, and a support team deployed one to triage tickets, each built on a different platform, reporting to a different owner. None of that required central approval, because none of it looked like a traditional IT project big enough to need one. The result, according to OutSystems' data, is that only 12 percent of organizations have implemented a centralized platform to manage the agents they already have, which means the other 88 percent are governing agentic AI through approaches that vary team by team and region by region, when they are governing it at all. For a leadership team trying to get ahead of that before it hardens into permanent shadow infrastructure, a documented AI governance framework gives every team building an agent the same starting rules instead of inventing its own.

Why confidence and control are moving in opposite directions

The more unusual finding is that sprawl hasn't dented anyone's confidence. Forty-nine percent of the IT leaders surveyed describe their organization's agentic AI capabilities as advanced or expert, even as 94 percent worry about what that capability has already produced. That combination, high self-assessed skill paired with low centralized control, tracks with research elsewhere showing executives tend to overestimate how far along their AI programs actually are relative to what the data supports. Separate research from EXL on UK business leaders found a similar gap between perceived and actual AI maturity, and the OutSystems numbers suggest the same overconfidence extends specifically to how well agent sprawl is being managed.

For an operating model still catching up to how many agents are already running inside it, the fix is rarely a mandate to stop building. It's visibility: an inventory of every agent in production, who owns it, what systems it touches, and what happens when it fails. Enterprises deciding where to start that inventory can use an AI agent library of vetted use cases as a reference point for what a well-governed agent actually looks like, rather than trying to retrofit governance onto whatever already shipped.

Key Takeaways

OutSystems' 2026 State of AI Development report surveyed nearly 1,900 global IT leaders worldwide and found that 96% are already using AI agents in some capacity, while 97% are exploring agentic AI strategies that span the whole enterprise.
Despite that adoption, 94% of the same IT leaders said agent sprawl, agents built by different teams with no shared oversight, is increasing complexity, technical debt, and security risk, and only 12% have a centralized platform in place to manage it.
Confidence hasn't caught up to that risk: 49% of IT leaders describe their organization's agentic AI capabilities as advanced or expert, even as the large majority worry about what their own organization has already built without central control.