Agentforce Execution in the Physical World

Proven since 2016

How Salesforce Agentforce decisions are executed in real environments using Cobalt’s security automation platform.

At NRF 2026, Salesforce and Cobalt showed what it means for AI agents to move beyond software.

Salesforce Agentforce and Voice Agent provide the communication and orchestration layer — whether that’s through Slack, voice, or interfaces like the Cobalt robot.

Cobalt Monitoring Intelligence is the system that understands what is happening in the physical environment and determines what should happen, when, and where. It then coordinates execution across physical systems and, when appropriate, uses the Cobalt robot as one of the ways actions are carried out.

This approach is not about robotics. It’s about giving enterprises a reliable way to operate AI in the real world.

Cobalt robot operating in a Salesforce office environment, demonstrating Agentforce execution in the physical world.

Why execution is the missing piece in enterprise AI

Most enterprise AI today still stops at recommendations, dashboards, and alerts.

But real value starts when AI can run parts of the business — especially in physical environments, where systems are fragmented, conditions change, and the cost of mistakes is high.

This is where execution becomes the hard problem.

Cobalt was built for exactly these kinds of environments. For years, it has operated in complex physical security operations where uptime, traceability, and reliability are non-negotiable. That same foundation now enables AI agents like Agentforce to move from deciding what should happen to actually making it happen.

What Salesforce and Cobalt demonstrated at NRF 2026

At NRF, Salesforce presented Agentforce and Voice Agent as the communication and orchestration layer — bringing Salesforce systems together and providing ways for people to interact with AI through tools like Slack, voice, and interfaces like the Cobalt robot.

Cobalt demonstrated how physical environments are actually operated: how context is understood, how decisions are made about what should happen, when, and where, and how those actions are carried out across real systems using existing infrastructure — without rip-and-replace and without relying on people to manually coordinate every step.

In this demonstration:

  • Agentforce and Voice Agent provided the communication and orchestration layer
  • Cobalt Monitoring Intelligence determined what should happen in the physical world and coordinated execution
  • The Cobalt robot served as one of the physical interfaces to make that process visible

The scenario itself is illustrative. The underlying Cobalt Monitoring Intelligence platform is what enterprises use today to automate and operate physical security and safety environments at scale.

Execution in the real world

This video shows Cobalt’s execution platform operating in a real environment, integrated with Salesforce.

While the interface includes a robot, the core capability is infrastructure-level: securely connecting systems, coordinating actions, and ensuring AI-driven decisions are carried out consistently in physical spaces.

This same execution layer is used today across enterprise security operations.

What this could mean for your organization

If you’re responsible for automation, AI programs, or enterprise systems, this approach enables:

  • AI agents that move from insight to action
  • Execution across physical environments using existing systems
  • Less manual coordination, fewer handoffs, and more consistent outcomes
  • A trusted foundation for operating AI in real-world conditions

FAQ

Is this a robot product or a robotics solution?

No. The robot is used as a physical interface to make agentic systems visible in the real world. Cobalt’s solution is an agentic SAAS platform that connects physical systems and executes actions across real environments, primarily in security and safety operations increasing efficiency and accelarating judgment.

Will Cobalt’s robot perform the retail workflows shown in the NRF demo?

No. The scenario shown at NRF is illustrative.

It demonstrates how agentic systems like Salesforce Agentforce can reason and how Cobalt’s platform can execute decisions in physical environments. Cobalt’s real-world deployments focus on security, safety, and physical operations — not customer-facing retail workflows.

What exactly does Salesforce Agentforce do in this setup?

Agentforce and Voice Agent handle communication and orchestration inside Salesforce. Cobalt Monitoring Intelligence determines what should happen in the physical world and coordinates execution. The Cobalt robot is one of the ways those actions are carried out.

What exactly does Cobalt do?

Cobalt Monitoring Intelligence is the system that understands what is happening in the physical environment and decides what should happen next, when, and where. It connects to systems like cameras, access control, sensors, and alarms, and applies rules, context, and automation to coordinate real-world actions.

The Cobalt robot is one of the ways those actions can be carried out. It executes tasks based on decisions made by Cobalt Monitoring Intelligence — not directly by Salesforce Agentforce.

Agentforce and Voice Agent act as communication and orchestration layers, allowing people to interact with the system through tools like Slack, voice, or interfaces like the robot.

Where is this used today?

Cobalt is used today in large enterprise environments to operate and automate physical security and safety operations.

The same execution platform shown here is already running in real-world, always-on environments where reliability and traceability matter.

Is this only relevant for retail?

No.

The NRF demonstration uses a retail setting because of the event context, but the underlying execution platform is designed for any enterprise with physical operations — including corporate campuses, critical infrastructure, manufacturing, logistics, and other complex environments.

Do I need to replace my existing systems to use this?

No.

Cobalt is designed to work with the systems organizations already have. It connects to existing physical security and operational systems and coordinates them, rather than requiring a rip-and-replace approach.

How is this different from dashboards or alerting systems?

Dashboards and alerts still rely on people to notice issues, decide what to do, and coordinate a response.

Cobalt’s platform focuses on execution. It allows decisions — whether made by humans or AI agents like Agentforce — to be carried out automatically and consistently in physical environments.

Is this available today?

Yes.

Cobalt has been running AI-driven automation and agentic workflows in real-world, always-on environments since 2016.

The platform shown here is not experimental. It is the same Cobalt Monitoring Intelligence that enterprises use today to operate and automate physical security and safety environments at scale — handling real incidents, coordinating real systems, and executing real actions every day.

What you see in the Salesforce demonstration builds on nearly a decade of production use in complex physical environments where reliability, traceability, and uptime matter.

Cobalt Robot patrolling in office general area

Request a Demo

Complete the form and one of our experts will reach out soon!

Demo request
Cobalt uses your contact information to communicate about our products and services. You can unsubscribe from these communications at any time. For details on how to unsubscribe and insights into our privacy practices and commitment to safeguarding your privacy, please refer to our Privacy Policy.
Cobalt Monitoring Intelligence , AI-powered, Human Verified monitoring solutions for Soc and Guardforce

Request a Demo

Complete the form and we will be in touch to set up a meeting!

Demo request
Cobalt uses your contact information to communicate about our products and services. You can unsubscribe from these communications at any time. For details on how to unsubscribe and insights into our privacy practices and commitment to safeguarding your privacy, please refer to our Privacy Policy.