Salesforce has expanded Agent Fabric – its multi-agent management layer – with a new centralised control plane that lets enterprises discover, govern, and orchestrate AI agents from any vendor in one place. The update is designed to solve a problem that has quietly become one of the biggest headaches in enterprise AI: too many agents, running on too many platforms, with no single view of what they’re doing or what they’re costing.
What Is Agent Fabric, and What Does This Expansion Actually Do?
Think of Agent Fabric as a command centre for AI agents. Most large enterprises today aren’t running a single AI system – they’re running dozens, often hundreds, of agents built on different platforms: Salesforce’s own Agentforce, Amazon Bedrock, Microsoft Foundry, OpenAI, homegrown models, and more. Until now, managing all of those from a central point has been close to impossible.
Agent Fabric, which launched in September 2025, was Salesforce’s first answer to that problem. This week’s expansion makes it significantly more capable – and more directly useful for organisations trying to scale AI without losing control of costs, compliance, or customer outcomes.
The core of the update is a new centralised control plane: a single layer that sits across an enterprise’s entire AI stack and handles four things – discovering what agents exist, connecting them to workflows, governing how they behave, and monitoring what they do.
Related Articles
The Enterprise Buyer’s Guide to Customer Journey Orchestration Platforms
AI in Customer Engagement: 2026 Insights
Why Personalization Is Driving Fatigue Instead of Engagement
What’s New, in Plain Terms
- Automatic agent discovery – Previously, IT teams had to manually register AI agents and tools across their environment. The expanded Agent Scanners now do this automatically – scanning across Amazon Bedrock, Microsoft Foundry, GoDaddy, and MCP servers, and cataloguing everything in a central Agent Registry. If an AI agent exists somewhere in your enterprise, Agent Fabric now finds it.
- A visual tool for building multi-agent workflows – A new drag-and-drop Visual Authoring Canvas lets developers map out how agents hand off tasks to each other — and where a human needs to step in — without heavy coding. It’s designed to shorten the time from “we want to deploy this AI workflow” to “it’s live in production.”
- Rules-based guardrails for autonomous agents – One of the trickiest problems with AI agents in customer-facing roles is knowing when to let them run freely and when to enforce a fixed process. A new feature called Agent Script for Agent Broker addresses this directly: it lets organisations define hard rules for critical handoffs, such as escalating a complaint or processing a refund, while still allowing the AI to reason and respond naturally in between. Salesforce calls this “guided determinism.”
- Central cost and compliance controls for LLMs – Enterprises running multiple large language models – OpenAI, Gemini, Salesforce’s own – have had no single point to manage token usage, enforce routing rules, or track spend. A new LLM Governance layer on AI Gateway fixes that. It also introduces Trusted Agent Identity, which requires mobile approval for high-stakes agent actions like financial transactions or legal review, creating an auditable trail for every privileged operation.
What Customers Are Saying
Diabsolut, a Salesforce implementation partner, is already running the platform across a network of agents that spans Agentforce, SharePoint, Certinia, and custom-built tools – all triggered from Slack.
John Pettifor, SVP of Innovation at Diabsolut, said:
“Agent Fabric lets us deploy a coordinated network of agents, quickly and safely. It provides a single governed, agentic layer for our implementation solution. Now, tasks that previously took days are completed in seconds. Agent Fabric is how we scale AI without losing control.”
Other named customers include Capita – one of the UK’s largest business process outsourcing firms and a major player in customer service delivery – and medical device company Alcon.
The Bigger Picture
Speaking at the launch of Agentforce 360 in October 2025, Salesforce CEO Marc Benioff set out the direction of travel:
“We’re entering the age of the Agentic Enterprise – where AI elevates human potential like never before.”
This week’s Agent Fabric expansion is the infrastructure that vision runs on. That ambition only works if the agents involved can be trusted – and trust, at enterprise scale, requires governance infrastructure. That’s the gap Agent Fabric is designed to fill.
The vendor-neutral positioning is deliberate and strategically significant. By building discovery and governance tools that work across Amazon, Microsoft, Google, and others, rather than only within Salesforce’s own ecosystem, the company is making a clear bet: enterprises will not standardise on a single AI vendor, and the winner in the long run will be whoever becomes the coordination layer across all of them.
For CX leaders specifically, the implications are direct. AI agents are increasingly handling the first, sometimes only, interaction a customer has with a brand. Without a control plane that can enforce consistent behaviour, flag failures, and manage costs across a fragmented AI estate, scaling that to enterprise level is a governance risk, not just a technical one.
Agent Fabric is available now, with the new capabilities rolling out globally. The platform has expanded to Canada and Japan and now supports on-premises and private cloud deployment for regulated industries.
What Comes Next
The Agent Fabric expansion signals where enterprise AI infrastructure is heading: away from isolated tools and towards governed, interoperable networks of agents that span every system a business runs.
In the near term, the pressure on CX teams will be to get their own AI estate in order – auditing what agents they have, what they’re doing, and whether they’re operating within acceptable risk and cost parameters. The organisations that do this now will be better placed to scale when the next wave of agentic capability arrives.
Longer term, the control plane model Salesforce is building points to a world where AI agents are not just executing tasks, but continuously learning from outcomes, adjusting their behaviour, and routing themselves to the most effective model for each job – all without human intervention. That future is closer than it looks. The governance infrastructure to support it, however, is still being built – and Agent Fabric is one of the first serious attempts to lay the foundations.
For CX professionals, the question is no longer whether AI agents will become central to how customers are served. It’s whether the systems managing those agents are mature enough to be trusted with that responsibility.
Frequently Asked Questions
What is Salesforce Agent Fabric?
Salesforce Agent Fabric is an enterprise AI orchestration platform that acts as a centralised control plane for managing AI agents.
The tool allows organisations to discover, connect, govern, and monitor AI agents across multiple vendors and platforms – including Salesforce Agentforce, Amazon Bedrock, Microsoft Foundry, OpenAI, and Gemini – from a single interface.
What did Salesforce announce about Agent Fabric in April 2026?
Salesforce announced a major expansion of Agent Fabric, adding four new capabilities: automated discovery of AI agents across third-party platforms, a visual drag-and-drop workflow authoring canvas, rules-based guardrails for multi-agent orchestration (called “guided determinism”), and a centralised LLM governance layer for managing costs, compliance, and routing across multiple large language models.
What problem does Salesforce Agent Fabric solve?
Agent Fabric addresses the growing challenge of AI agent sprawl – where enterprises are running AI tools from multiple vendors with no unified way to manage, monitor, or govern them. It gives IT and CX teams a single point of oversight for every agent interaction, helping control costs, maintain compliance, and ensure consistent customer outcomes.
What is “guided determinism” in Salesforce Agent Fabric?
Guided determinism is Salesforce’s approach to balancing AI autonomy with business control. It allows organisations to define fixed rules for critical workflow steps – such as how a complaint is escalated or a refund is processed, while still allowing AI agents to reason and respond flexibly in between those checkpoints. This makes autonomous agents safer to deploy in regulated or high-stakes customer-facing environments.
Is Salesforce Agent Fabric vendor-neutral?
Yes. Agent Fabric is designed to work across multiple AI vendors and platforms, not just Salesforce’s own products. Its Agent Scanners can automatically discover and register agents running on Amazon Bedrock, Microsoft Foundry, GoDaddy, and MCP servers. Salesforce’s LLM Governance layer also supports models from OpenAI and Google Gemini alongside its own.
How does Agent Fabric affect customer experience teams?
For CX leaders, Agent Fabric provides the governance infrastructure needed to deploy AI agents in customer-facing roles at scale. It ensures that agents behave consistently, that high-risk actions (such as financial transactions) require human approval, and that costs are managed across the entire AI stack – reducing the risk of deploying autonomous agents in live customer journeys.
Which companies are using Salesforce Agent Fabric?
Salesforce has named Capita, Alcon, and Diabsolut as customers already managing and orchestrating agents across diverse AI platforms using Agent Fabric. Since its launch in September 2025, the platform has managed thousands of agentic instances across enterprise deployments.
Where is Salesforce Agent Fabric available?
Agent Fabric is available globally, with recent expansion to Canada and Japan. It also now supports Runtime Fabric deployment, allowing organisations to run governance guardrails directly on their own private cloud or on-premises infrastructure – an important feature for regulated industries with strict data residency requirements.
