Salesforce has announced MuleSoft Agent Fabric, a new solution designed to solve one of the most pressing challenges of enterprise AI adoption: governing and orchestrating AI agents that are proliferating across organizations without centralized management.
Agent Fabric provides a single place to register, orchestrate, govern, and observe every AI agent, regardless of where it was built — whether on Salesforce's own Agentforce platform, third-party tools, or custom internal systems.
The Enterprise Agent Sprawl Problem
As enterprises deploy AI agents across departments and functions, a governance challenge has emerged: multiple teams building and deploying agents independently, without coordination or centralized oversight. This creates:
- Security gaps: Agents accessing systems and data without consistent access controls
- Redundancy: Multiple agents performing similar functions across different departments
- Observability blind spots: No single view of what agents are doing, when, or how well
- Compliance risk: Agents making decisions or accessing data without audit trails
- Integration conflicts: Agents competing for the same system resources or producing conflicting outputs
Agent Fabric addresses these challenges by providing an infrastructure layer that sits above individual agents and below enterprise governance.
How Agent Fabric Works
The solution extends MuleSoft's existing API and integration infrastructure into the AI agent domain:
Centralized registry: Every AI agent, regardless of its platform or origin, is registered in a single catalog that tracks its capabilities, permissions, status, and performance metrics.
Orchestration: Agent Fabric coordinates multi-agent workflows, routing tasks to the right agent based on capability, availability, and policy rules. This enables complex processes that span multiple agents and systems.
Governance: Policy enforcement is applied consistently across all agents, including access controls, data handling rules, and decision boundaries. Administrators can define what agents can and cannot do, regardless of which team built them.
Observability: A unified dashboard provides real-time visibility into agent activity, performance, and anomalies across the entire agent network.
Agentforce and MuleSoft Integration
Agent Fabric builds on the existing Agentforce-MuleSoft integration that connects AI agents to enterprise systems via APIs:
- Developers can create topics with actions and instructions from APIs, enabling agents to interact with systems like NetSuite, Jira, SAP, and Workday
- Administrators can launch multi-system agents faster using pre-built connectors
- Business teams can build no-code workflows that enable agents to take action across multiple systems
The practical result: AI agents can automate everyday workforce tasks like leave requests, timesheet logging, expense processing, and onboarding workflows — tasks that previously required employees to navigate multiple systems manually.
Real-World Deployment
Salesforce itself is using the platform internally. The company uses MuleSoft to integrate employee data across 90+ internal and external systems, and has implemented Agentforce with these integrations to:
- Understand employee needs proactively
- Resolve data discrepancies with minimal human involvement
- Enhance the overall employee experience through automated support
- Connect HR, finance, and operations workflows through a unified agent layer
Market Context
The launch of Agent Fabric reflects a broader industry trend: as AI agents move from pilot to production, the infrastructure to manage them is becoming critical.
Deloitte's 2026 Tech Trends report found that only 11% of enterprises run agentic AI in production, with governance and observability cited as primary barriers. Products like Agent Fabric directly address these barriers.
Gartner predicts that 33% of enterprise software applications will include agentic AI by 2028, up from less than 1% today. Managing this proliferation requires exactly the kind of orchestration layer that Agent Fabric provides.
Implications for Virtual Assistant and Outsourcing Services
The maturation of enterprise AI agent infrastructure has direct implications for virtual assistant providers:
New oversight roles: As enterprises deploy agent networks managed through platforms like Agent Fabric, they need human operators to monitor agent performance, handle escalations, and manage exceptions. Virtual assistants trained in AI agent supervision can fill these emerging roles.
Process design opportunity: The shift from manual processes to agent-orchestrated workflows requires process redesign. Service providers who understand both human operations and AI agent capabilities can help companies design effective hybrid workflows.
Integration support: Connecting agents to enterprise systems through APIs and MuleSoft connectors requires technical coordination that outsourced teams can manage — from initial configuration to ongoing maintenance.
SMB accessibility gap: While Agent Fabric targets enterprise-scale deployments, small and mid-size businesses need similar capabilities at their scale. VA service providers can deliver AI-augmented operations to SMBs who can't invest in enterprise agent infrastructure but need the productivity benefits.
The launch of Agent Fabric signals that enterprise AI has moved past the question of "should we deploy agents?" to "how do we manage the agents we've already deployed?" The infrastructure layer is catching up to the application layer — and the services layer needs to keep pace.