Enterprise networks are becoming filled with AI agents, creating governance blind spots for leaders managing multicloud infrastructure.
As various business units race to deploy generation technologies, CIOs especially find themselves with fragmented and unmonitored assets in their ecosystems. This reflects the challenges of shadow IT in the cloud era, but involves autonomous actors who can execute business logic and access sensitive data.
IDC predicts that the number of actively deployed AI agents will exceed 1 billion by 2029. This is an increase of 40 times the current level. In the first half of 2025 alone, agent creation jumped 119%. The immediate challenge for enterprise leaders will shift from building these agents to finding, auditing, and managing them across platforms.
Salesforce addressed this fragmentation by extending the capabilities of MuleSoft Agent Fabric and introducing an auto-discovery tool designed to centrally manage AI agents, regardless of their origin.
Automated discovery
Visibility remains a core issue for security and operations teams. When marketing teams deploy AI agents on one platform and logistics teams build on another, effective governance becomes difficult because central IT loses a unified view of the organization’s digital workforce.
MuleSoft’s updated architecture addresses this issue through an “agent scanner.” These tools continuously patrol major ecosystems such as Salesforce Agentforce, Amazon Bedrock, and Google Vertex AI to identify running agents. Rather than having developers manually register deployments, the system automates discovery.
Finding an agent is only the first step. Compliance leaders need to understand the logic behind it. The scanner extracts metadata detailing the agent’s capabilities, the LLM that powers it, and the specific data endpoints it is allowed to access. This information is normalized into standard agent-to-agent (A2A) specifications to create a uniform profile of assets regardless of the underlying vendor.
“The most successful organizations over the next decade will be those that take full advantage of the diversity of multicloud AI environments,” said Andrew Comstock, SVP and GM, MuleSoft. “The expanded capabilities of MuleSoft Agent Fabric give them the freedom to innovate across any platform while maintaining the unified visibility and control they need to scale.”
AI agent governance and cost management
Unmanaged agents are exposed to financial inefficiencies and risks. Consider a CISO in the banking sector. In standard operations, validating a new loan processing agent requires manual documentation tracking from the development team. Automated cataloging allows security teams to instantly see which financial databases agents are accessing and check their authorization levels without manual intervention. This feature allows security teams to see real-time data instead of old snapshots.
From a financial perspective, visibility drives integration. Large companies often suffer from redundancy as regional teams source or build similar tools on their own. For example, a multinational manufacturer may have three separate teams paying separate aggregation agents on different platforms.
By using MuleSoft Agent Visualizer to filter assets by job type, operations leaders can identify these duplicates. Consolidating these into one high-performance asset reduces excess licensing costs and allows you to reallocate your budget toward new development.
Successfully transitioned to an “agency company”
Innovation often happens at the edge, where data scientists build bespoke tools outside of formal procurement channels.
The Enhanced Agent Fabric addresses this issue by allowing you to register “homegrown” agents and Model Context Protocol (MCP) servers via a URL. This is particularly relevant in areas such as logistics, where teams may build their own internal tools for database optimization. Rather than leaving these assets hidden, you can register them to make them discoverable and reusable across your company.
Jonathan Harvey, Head of AI Operations at Capita, said, “Agent Scanner allows us to focus on innovation instead of inventory management. Knowing that all agents are automatically discovered and cataloged allows our teams to collaborate, reuse work, and build smarter multi-agent solutions.”
Similarly, AT&T leverages this framework to coordinate agents across customer support, chat, and voice interactions.
Brad Ringer, Enterprise and Integration Architect at AT&T, explained, “With rapid advances in AI, MuleSoft Agent Fabric provides the framework we need to scale. It helps us integrate and orchestrate all the agents and MCP servers we’re building for customer support, chat, and voice interactions. It’s more than just a tool; it’s a big enabler for everything we do next.”
The move to an “agent enterprise” requires a change in governance around how IT assets are tracked, and the days of managing integrations via old spreadsheets are no longer compatible with the speed of AI agent deployment.
Leaders should assume that their AI agent inventory is incomplete and deploy automated scanning tools to establish a baseline of truth. Once this baseline is established, governance policies should require that all agents, whether purchased or built, expose their capabilities and data access privileges in a standardized format, such as A2A, to facilitate monitoring.
Finally, executives can use the visibility these tools provide to audit spending, identify duplicate functionality across cloud environments, and consolidate them to control total cost of ownership (TCO).
As organizations move from pilot programs to large-scale deployments, the differentiator is not the intelligence of individual agents, but the consistency of the network that connects them.
See: Balancing AI cost efficiency and data sovereignty
Want to learn more about AI and big data from industry leaders? Check out the AI
AI News is brought to you by TechForge Media. Learn about other upcoming enterprise technology events and webinars.
