Controlling AI Agent Sprawl: A CIO’s Essential Guide to Governance Strategies

Corporate networks are increasingly populated with AI agents, leading to significant governance challenges for leaders overseeing multi-cloud infrastructures. As various business units quickly embrace generative technologies, CIOs find their ecosystems cluttered with fragmented and inadequately monitored assets, reminiscent of the shadow IT complications that arose during the cloud era. The emerging issue highlights the presence of autonomous actors capable of executing business logic and gaining access to sensitive data.

According to IDC, the number of actively deployed AI agents is expected to surpass one billion by 2029, which marks an astounding forty-fold increase from current levels. In just the first half of 2025, there was a staggering 119 percent surge in the creation of these agents. Consequently, enterprise leaders are challenged not only to create these agents but also to locate, audit, and govern them effectively across various platforms.

In response to this growing fragmentation, Salesforce has enhanced its MuleSoft Agent Fabric capabilities with the introduction of automated discovery tools designed to centralize the management of AI agents, regardless of their origins.

Automating Discovery

Visibility is a critical concern for security and operations teams. When marketing and logistics teams deploy AI agents on different platforms, effective governance becomes difficult, and central IT loses a consolidated view of the organization’s digital workforce. To tackle this, MuleSoft’s newly upgraded architecture utilizes "Agent Scanners." These continuously monitor major ecosystems, including Salesforce Agentforce, Amazon Bedrock, and Google Vertex AI, to identify active agents automatically. This process eliminates the tedious need for developers to register their deployments manually, facilitating easier oversight.

However, locating an agent is only the initial step; compliance teams must also grasp the logic behind each agent. The scanners gather metadata that outlines an agent’s capabilities, the underlying large language models (LLMs) driving its operations, and the specific data endpoints it is authenticated to access. This information is standardized into Agent-to-Agent (A2A) specifications, forging a uniform profile for assets across different vendors.

According to Andrew Comstock, SVP and GM of MuleSoft, the organizations that embrace the diversity within the multi-cloud AI landscape will emerge as the most successful in the coming decade. MuleSoft’s enhanced capabilities promote innovation across any platform while ensuring the necessary unified visibility and control required for scaling.

Governance and Cost Control for AI Agents

Failing to manage these agents could lead to financial inefficiencies and increased risk exposure. For instance, a Chief Information Security Officer (CISO) in the banking sector would typically need to manually chase documentation from development teams to verify a new loan-processing agent. However, automated cataloging enables security teams to instantly access information related to which financial databases an agent interacts with and confirm its authorization levels without manual intervention. This ensures that security teams have real-time visibility into operations rather than relying on outdated data.

From a budgetary perspective, heightened visibility fosters consolidation. Large enterprises often encounter redundancies, where regional teams independently procure or develop similar tools. For example, a multinational manufacturer may have multiple teams purchasing separate summarization agents across different platforms. By employing the MuleSoft Agent Visualizer, leaders can filter the operational landscape by job type, helping them identify overlaps. This allows for consolidating redundant assets, lowering licensing costs, and reallocating budgets toward innovative development.

Transitioning Successfully to an ‘Agentic Enterprise’

Innovation frequently emerges at the periphery, where data scientists create custom tools outside of formal procurement channels. The expanded Agent Fabric accommodates this by enabling the registration of “homegrown” agents and Model Context Protocol (MCP) servers via URL. This is especially beneficial in sectors such as logistics, where internal tools for proprietary database optimization may be developed. By registering these assets, companies can unlock their potential for reuse across the organization.

Jonathan Harvey, Head of AI Operations at Capita, noted that Agent Scanners would shift the focus from inventory management to innovation. Knowing that every agent is cataloged automatically facilitates collaboration and reuse of creations, allowing for smarter multi-agent solutions.

Similarly, AT&T is tapping into this framework to orchestrate agents across customer support, chat, and voice interactions. Brad Ringer, Enterprise & Integration Architect at AT&T, emphasized that MuleSoft Agent Fabric is not just a tool but a significant enabler for future initiatives.

Transitioning into an "Agentic Enterprise" necessitates a shift in governance habits regarding asset tracking, making it imperative to abandon outdated methods like managing integrations through old spreadsheets. Leaders should assume their inventory of AI agents is incomplete and deploy scanning tools to establish an accurate baseline. Once this baseline is confirmed, governance policies must mandate that all agents, whether developed internally or sourced externally, disclose their capabilities and data access privileges in a standardized format, such as A2A, to streamline monitoring.

Leveraging the insights offered by these tools, executives can conduct comprehensive audits of spending, uncovering duplicate functionalities spread across cloud environments and integrating them to reduce the Total Cost of Ownership (TCO).

As organizations evolve from isolated pilot programs to comprehensive deployments, success will hinge not on the intelligence of individual agents but on the cohesive network that interlinks them.

Discover the pinnacle of WordPress auto blogging technology with AutomationTools.AI. Harnessing the power of cutting-edge AI algorithms, AutomationTools.AI emerges as the foremost solution for effortlessly curating content from RSS feeds directly to your WordPress platform. Say goodbye to manual content curation and hello to seamless automation, as this innovative tool streamlines the process, saving you time and effort. Stay ahead of the curve in content management and elevate your WordPress website with AutomationTools.AI—the ultimate choice for efficient, dynamic, and hassle-free auto blogging. Learn More