Enterprise IT leaders face an uphill battle: Intricate data flows, legacy systems, and constant integration demands strain resources and slow innovation. For CIOs, the pressure to boost efficiency and productivity has never been greater. In fact, 86% of IT leaders expect workloads to rise in coming years, a Salesforce survey found.
Fortunately, there’s a compelling solution: AI agents operating autonomously behind the scenes — collecting data, reasoning, making decisions, and taking action. Unlike AI assistants or copilots that require direct user interaction, these so-called ambient or headless agents “wake up” when triggered by data, workflows, APIs, or even other agents and then run independently, 24/7.
Having teams of ambient agents operating in the background enables CIOs to automate entire swaths of static, hand-crafted workflows, streamlining operations and building more dynamic, flexible, and scalable IT environments.
Jayesh Govindarajan, Executive Vice President of AI, Salesforce
Having teams of ambient agents operating in the background enables CIOs to automate entire swaths of static, hand-crafted workflows, streamlining operations and building more dynamic, flexible, and scalable IT environments. What’s more, by streamlining multistep processes and integrating disparate systems, they can substantially reduce manual workloads and costs. The effect: IT leaders gain more time to focus on strategic innovation and operational excellence. At the same time, employees become more productive as they are relieved from time-wasting “swivel chair” activities, where they bounce between apps and data to complete tasks.
Redefining the agentic landscape
My colleague Adam Evans, EVP and GM of Salesforce AI, believes this trend toward agents operating in the background and interacting with apps on our behalf will result in agents becoming the primary UI we use instead of those apps. I agree.
How will we get there? It starts with a multi-agent architecture. After the initial user prompt happens on the front end, an orchestrator agent in the background will manage task delegation and workflow coordination, ensuring user requests are routed to the right systems. Concurrently, specialized ambient agents will activate, analyze the orchestrator’s input, and engage other agents and back-end systems to gather whatever data is needed to reason, respond, and take action.
Here’s how that might work in a Salesforce environment. A customer goes online to ask about an order’s status. The system automatically activates a series of specialized ambient agents — no human involvement needed. A service agent then retrieves order details from Service Cloud, a commerce agent checks shipment status in Commerce Cloud, and a marketing agent pulls data and metadata from Data Cloud, the hyperscale data engine for the deeply unified Salesforce Platform. Finally, an orchestrator agent synthesizes all of the information to accurately answer the customer’s inquiry.
Because agents can access both data and metadata, they also gain a deeper understanding of the customer, allowing them to deliver highly personalized product and service recommendations.
Looking ahead to a multi-agent future
Today, thousands of companies like SharkNinja, Vivint, and the Adecco Group are deploying custom agents they’ve built with Agentforce, Salesforce’s digital labor platform. In fact, the numbers keep growing, with Salesforce signing more than 5,000 Agentforce deals, including more than 3,000 paid deals, in the fourth quarter of fiscal year 2025 alone.
It’s impressive momentum, but most of these agents still operate independently, handling specific tasks in isolation rather than working in tandem. Multi-agent orchestration involving behind-the-scenes agents is the next big step for agentic systems, and it’s coming — bet on it. Over time, multi-agent systems will function as seamlessly as biological processes, where specialized cells exchange signals to maintain our health.
Multi-agent orchestration involving behind-the-scenes agents is the next big step for agentic systems, and it’s coming — bet on it.
Jayesh Govindarajan, EVP of AI, Salesforce
Getting there will require overcoming a few technical hurdles. Just as cellular processes rely on reliable biochemical protocols to communicate with one another, multi-agent systems will need standardized methods for dynamic discovery and seamless collaboration. Industry cooperation — similar to how DNS enables internet domain discovery — will be key to advancing multi-agent architectures.
High-quality data and metadata are also essential, allowing orchestrator agents to understand each ambient agent’s capabilities and limitations. Solutions like Salesforce Data Cloud, therefore, will be critical. Finally, enterprises will require agentic systems to exist on a well-integrated foundation like the Salesforce Platform, where data, metadata, and actions exist in a single ecosystem, simplifying agent deployment and orchestration. Salesforce’s deeply unified platform is its unique advantage.
Build the future with Agentforce.
A unified multi-agent system
For IT leaders, adopting a multi-agent architecture is about making work more efficient and workers more productive. Ambient agents capable of working with one another to complete tasks and projects will have a profound effect on organizations — as long as they’re built on a solid platform, can access the right systems, apps, and data, and can find one another in industry-standard ways.
I’m optimistic we’ll see this soon because the pieces are falling into place.
Go deeper:
- Learn why agents will be the new UI in enterprise workflows
- Watch this video demo of headless agents
Just For You
Build and customize autonomous AI agents to support your employees and customers 24/7.
Just For You
Build and customize autonomous AI agents to support your employees and customers 24/7.
