Enterprises face an escalating challenge as artificial intelligence agents increasingly populate their digital infrastructures. This surge creates substantial oversight gaps, particularly for CIOs navigating complex multi-cloud environments. Much like the 'shadow IT' phenomenon of the cloud era, autonomous AI agents, capable of executing intricate business logic and accessing sensitive data, are being deployed by various departments without unified central visibility.
Industry analysts, including IDC, project a dramatic increase in actively deployed AI agents, estimating over one billion by 2029 – a forty-fold jump from current levels. Agent creation alone surged by 119 percent in the first half of 2025. This rapid expansion shifts the immediate focus for enterprise leadership from merely developing these agents to comprehensively locating, auditing, and effectively governing them across diverse platforms.
Automating Agent Discovery and Compliance
Salesforce has addressed this growing fragmentation by enhancing its MuleSoft Agent Fabric capabilities. This expansion introduces automated discovery tools designed to centralize AI agent management regardless of their deployment origin. A primary concern for security and operations personnel is the lack of comprehensive visibility. When different departments, such as marketing or logistics, independently deploy AI agents on various platforms, central IT struggles to maintain a unified overview of the organization's digital workforce, hindering effective governance.
MuleSoft’s refined architecture tackles this through ‘Agent Scanners’. These tools proactively monitor prominent AI ecosystems – including Salesforce Agentforce, Amazon Bedrock, and Google Vertex AI – to pinpoint active agents. This approach removes the burden on developers to manually register their deployments, automating the detection process. Beyond mere discovery, these scanners delve deeper, extracting crucial metadata. This data reveals an agent's specific functionalities, the large language models (LLMs) powering it, and the precise data access points it is authorized to utilize. This information is then standardized into Agent-to-Agent (A2A) specifications, creating a consistent profile for all AI assets, irrespective of their originating vendor.
Andrew Comstock, SVP and GM of MuleSoft, commented on this evolution: “The most successful organizations of the next decade will be those that harness the full diversity of the multi-cloud AI landscape. The expanded capabilities of MuleSoft Agent Fabric give you the freedom to innovate across any platform while maintaining the unified visibility and control needed to scale.”
Enhanced Governance and Cost Optimization
The proliferation of unmanaged AI agents poses significant financial inefficiencies and elevates risk exposure. For example, a Chief Information Security Officer (CISO) in the banking sector typically faces the arduous task of manually requesting documentation to verify a new loan-processing agent from development teams. Automated cataloging, however, enables security personnel to instantly ascertain which financial databases an agent interacts with and confirm its authorization levels, eliminating manual delays. This ensures security teams operate with current data, not historical snapshots.
Furthermore, enhanced visibility directly contributes to cost savings through consolidation. Large corporations often experience redundancy, with various regional teams independently acquiring or developing similar AI solutions. A global manufacturing firm, for instance, might find three distinct teams funding separate summarization agents across different platforms. Tools like the MuleSoft Agent Visualizer empower operations leaders to filter their agent inventory by function, pinpointing such overlaps. Consolidating these redundant functions into a single, optimized asset can drastically cut licensing expenditures and free up resources for new innovation.
Fostering an 'Agentic Enterprise' Through Innovation and Reuse
Innovation frequently originates from decentralized efforts, where data scientists develop specialized tools beyond established procurement processes. The expanded Agent Fabric facilitates this by enabling the registration of internally developed ('homegrown') agents and Model Context Protocol (MCP) servers via a simple URL. This functionality proves especially valuable in sectors like logistics, where teams might create proprietary internal tools for database optimization. Rather than remaining isolated, these assets can be registered, becoming discoverable and reusable throughout the entire organization.
Jonathan Harvey, Head of AI Operations at Capita, emphasized this benefit, stating that “Agent Scanners will let us focus on innovation instead of inventory management. Knowing that every agent is automatically discovered and catalogued allows our teams to collaborate, reuse work, and build smarter multi-agent solutions.” Likewise, AT&T leverages this framework to coordinate agents across diverse customer interaction points, including support, chat, and voice. Brad Ringer, Enterprise & Integration Architect at AT&T, explained that “MuleSoft Agent Fabric provides the framework... to scale,” noting its role in orchestrating agents and MCP servers built for customer engagement. He characterized it as “a huge enabler for everything... next.”
Strategic Imperatives for Leaders
Successfully transforming into an 'Agentic Enterprise' necessitates a fundamental shift in IT asset governance. Traditional methods, such as tracking integrations through outdated spreadsheets, are ill-suited for the rapid pace of AI agent deployment. Enterprise leaders should operate under the assumption that their current AI agent inventory is incomplete. Implementing automated scanning solutions is crucial to establish an accurate foundational record.
Following this, governance frameworks should mandate that all agents – whether procured or internally developed – disclose their functionalities and data access permissions in a standardized format, such as A2A, to simplify ongoing monitoring. Ultimately, executives can leverage the insights gained from these tools to rigorously audit expenditures, pinpointing redundant capabilities across various cloud environments. Consolidating these overlaps will effectively manage and reduce the total cost of ownership (TCO). As organizations transition from initial pilot projects to widespread deployment, their competitive edge will derive not from the individual intelligence of agents, but from the interconnectedness and coherence of their entire agent network.
This article is a rewritten summary based on publicly available reporting. For the original story, visit the source.
Source: AI News
