4 Ways Salesforce is Going Headless Internally (and How You Can Too)

Agents and humans are working closer together than ever before. 

But in order to build agents, apps, and AI experiences that your customers and employees can trust, you need more than a good model. You need a foundation that gives both people and agents the full picture to do their work.

That’s why we built AIforce. Powered by the Headless Toolkit, it brings all of Salesforce’s capabilities, enterprise-grade security, and your unique business context directly to any surface where work is happening. By moving away from a static software model, we are shifting toward a central infrastructure for work that happens everywhere. 

Our internal use cases, your headless blueprint 

At Salesforce, we deeply believe that we cannot just build tools for our customers without proving their value internally first. We call this principle “Customer Zero.” By testing our innovations in one of the most demanding enterprise environments in the world — our own — we ensure they are tested and ready for scale. 

In this blog, we share four ways we’ve used AIforce to support our employees and drive innovation for our customers. Dive into each use case and see how you can bring them to your own business. 

1. Shipping help agents to support our workforce, wherever they work

With over 80,000 global employees, we need solutions that scale and support our workforce wherever they are.

• For IT support: By deploying specialized IT support agents, human employees and digital agents co-work in real time on the exact same records and tasks. Employees can access the IT support agent wherever is most convenient for them, either right inside Slack or via our dedicated self-service portal. Built on Agentforce, the agent’s reasoning engine interprets each request, routes it to the right topic or sub-agent, and follows that topic’s instructions to resolve it. The same Flows, monitoring, and governance run underneath wherever employees want to reach it — we built the logic once and surfaced it in both places. This new intake process has slashed resolution time to near-instantaneous automated execution. 
• For security and permissions: This agent looks out for specific system access requests, rapidly cross-references our complex organizational governance rules, finds the absolute right access level for the request, and then automatically routes and manages the necessary approvals. Built as a Slackbot skill, it interprets the request and calls tools exposed by the Salesforce MCP server. This eliminates the tedious manual oversight typically required for system provisioning and ensures compliance is seamlessly maintained. 

How do we develop a truly headless, trusted agent? There are three baseline components we use here at Salesforce:

1. It must be grounded in our business context via Data 360 to understand requests and find the right access. 
2. It has the ability to interpret a request and take action using the appropriate tools.
3. Governance is built in because autonomy requires guardrails.

2. Letting customers play with build-once, run-anywhere experiences

Our very first application that we shipped on Multi-Framework, using React, unlocked a new way for our customers to learn and interact with our products. 

The Headless Experience Layer is how we can separate what an agent does from how it appears visually. It translates and renders governed logic natively on whichever surface the developer chooses. 

The Headless Experience Layer playground allows developers to get hands-on with the Headless Experience Layer, understand the different components and tiles that are available, and create rich widgets using mock data. And most of all, they can immediately see what those custom widgets look like as they render across various surfaces.

The playground became our ultimate proof point. Built on Multi-Framework, it showed that an industry-standard framework like React can run securely on Salesforce, serve real external traffic, and get security and user authentication handled automatically.

3. Monitoring our 15,000-engineer org with one React app

Our support and operations team is responsible for our own internal production Salesforce org, which serves our entire global workforce, with around 15,000 active engineers building and deploying customized solutions internally on a daily basis.

This scale requires a highly mature operation and support mechanism just to stay healthy and functional. Launching an app or maintaining an org is never a finish line. Real success requires ongoing observability to spot issues early and optimize over time. To effectively manage it, internal engineering teams built a Multi-Framework React app that aggregates everything about the holistic health of the org’s various systems into one unified place. It gives our engineering teams critical, real-time signals on vulnerabilities such as stale or completely unused data, background processes timing out, massive backlogs of pending requests, and support cases rapidly approaching their SLAs. By maintaining clear visibility through observability tools, our teams can actively avoid business disruptions and continuously monitor system behavior. 

4. Building innovation from our own products

The best test of a platform is whether its makers build on it. Our own product teams do, and the results speak for themselves. 

• Builder Central: Our premier no-code app and agent builder. In just a rapid, 2-week sprint, our engineering team rebuilt Builder Central’s entire complex frontend in React with Multi-Framework. To put that into perspective, executing that same massive UX overhaul would have predictably taken two to three months using our previous legacy stack.
• Agentic Mobile App Builder: This tools helps internal teams design native mobile applications using natural language and visual editing, publish them directly to the Apple App Store and Google Play, and then continuously measure user adoption after launch. The Agentic Mobile App Builder is, itself, a modern React application hosted directly inside the Salesforce org, taking full advantage of Multi-Framework.

The speed and innovation would not have been possible without fundamentally changing the way we build. A core pivot was how we addressed the “inner-loop,” which refers to the repetitive cycle a developer experiences when making a code change and waiting to see it rendered live: edit, save, preview, and repeat. Before this transformation, every small incremental change required a build process that could take five to ten minutes, system syncs of around thirty minutes, and massive database updates pushing forty-five minutes — all before a developer could even see their visual change. By using React that development loop now takes mere seconds. 

Engineering and design have also changed how they work together. Modern design tooling is heavily built with React in mind, so our UX designers can now write and contribute code directly into the exact same repository that our engineers use. The feedback loop between a great design idea and a shipped, usable feature is exponentially shortened. The combination of speed, quality, and a drastically smoother design-to-dev handoff is exactly what AIforce unlocks for any team building on the Salesforce Platform.

The power of a unified, trusted platform

The browser was an incredibly necessary interface for a world where humans did all of the heavy lifting. But in a world where autonomous agents can proactively help, the most powerful interface is an optional one.

For Agentic Enterprises evaluating where to invest in AI agents, the platform question matters just as much as the agent question. Your AI is only as capable as the deep data and logic it is allowed to access. Build and secure your agents, apps, and experiences on the Salesforce Platform.