The age of AI experimentation is over. According to new research from OutSystems, 96% of organisations are already using AI agents in some capacity, and 97% are actively exploring strategies for system-wide agentic AI deployment. Nearly half (49%) describe their agentic AI capabilities as advanced or expert.

Those are remarkable adoption numbers. They tell you that AI agents are no longer a pilot project for forward-looking enterprises. They are a live operational reality for most businesses.

The follow-up number, though, tells a different story.

Ninety-four percent of organisations report that AI sprawl is increasing complexity, technical debt, and security risk.

The enterprise world has rushed into agentic AI, and the governance frameworks needed to manage it have not kept pace.

What AI Sprawl Actually Looks Like

The OutSystems 2026 State of AI Development report, compiled from nearly 1,900 global IT leaders surveyed between December 2025 and January 2026, paints a picture of fragmented adoption. Thirty-eight percent of organisations globally are running a mix of custom-built and pre-built agents, creating AI stacks that are difficult to standardise, monitor, or secure.

Different teams are building different agents. Different regions are deploying different tools. Individual departments are selecting AI vendors based on local needs without a centralised view of what is running, what data it has access to, or what it is actually doing.

This is AI sprawl. And only 12% of organisations have implemented a centralised platform to manage it. The remaining 88% are governing their AI agents with approaches that vary by team and region, or not really governing them at all.

Why This Matters More Than the Headline Numbers

The adoption numbers are impressive. The governance gap is the real story.

When AI agents are running across a business without centralised oversight, several problems compound quickly.

Security exposure increases. AI agents often have access to sensitive business data, customer records, or internal systems. When dozens of agents are running across fragmented environments (built by different teams, connected to different data sources, with different access controls) the attack surface expands rapidly. Any one of them can become an entry point.

Technical debt accumulates. Custom-built agents that solve a specific team’s problem often do not integrate cleanly with the rest of the organisation’s systems. Over time, you end up maintaining a portfolio of incompatible AI tools rather than a coherent infrastructure.

Accountability becomes unclear. When an AI agent makes a consequential decision (approving a transaction, escalating a customer complaint, routing a workflow) someone needs to be responsible for that outcome. In fragmented agentic environments, that accountability is often undefined.

Costs are hard to control. Without a central view of AI deployments, it is difficult to track spending, identify redundant tools, or negotiate effectively with vendors. Multiple teams paying separately for overlapping capabilities is common and expensive.

The Governance Gap Is a Strategic Risk

Gartner projects that 40% of enterprise applications will include task-specific AI agents by the end of 2026. That trajectory makes the current governance gap more urgent, not less. The organisations building AI agent infrastructure now, without centralised oversight, are creating a problem that will be significantly harder to untangle at scale.

The 12% of companies with centralised governance platforms have a structural advantage over the 88% that do not. They can audit what their agents are doing, respond faster to failures, and scale new use cases from a stable foundation.

The 94% who are worried about sprawl are right to be worried. Concern is appropriate. But concern without action just means the sprawl continues while you think about it.

What Organisations Should Do Now

The companies that OutSystems found in the most effective position are not necessarily the ones with the most agents. They are the ones that made governance a first-class concern early enough in their agentic journey that it shaped their deployment decisions from the start.

For organisations already in sprawl, the path forward typically involves three moves:

Audit what you have. Before you can govern AI agents effectively, you need a complete picture of where they are running, what data they can access, and what decisions they are influencing. Most organisations do not have this view.

Define ownership. Each AI agent should have a named owner who is accountable for its performance, its security posture, and its alignment with business objectives. Ownership without accountability is just bureaucracy.

Standardise incrementally. You do not need to tear out everything and start again. The goal is to gradually migrate fragmented deployments toward a common governance framework, starting with the highest-risk agents first.

What This Means for Business

The OutSystems research captures a moment that every business leader operating in 2026 recognises: AI agents are everywhere, and the processes for managing them are lagging behind.

The 6% of organisations not yet using AI agents will soon be in a very small minority. The question is not whether to adopt agentic AI. That decision has already been made by competitive pressure. The question is whether your deployment will be governed effectively enough to be sustainable.

The 94% who are worried about sprawl have already identified the problem. What differentiates them going forward is whether that concern turns into a structured governance approach before the complexity becomes unmanageable.

---

If you’re building with Claude or Codex right now, grab the free Working With Claude field guide. Thirty-two pages on the full ecosystem, Claude Code in depth, and how to roll agents out properly. Get the free guide.