How are MuleSoft and Salesforce combining integration, automation, and AI into an automation stack that delivers “entirely new enterprise superpowers”?
Integration, automation, and AI are converging into a single enterprise automation stack rather than remaining separate disciplines. At Salesforce’s Agentforce Tour in New York City on December 10, MuleSoft and Salesforce framed APIs, RPA, and AI agents as the building blocks of “entirely new enterprise superpowers” when designed and operated together. The future stack is unified, metadata-driven, and capable of orchestrating intelligent workflows across both legacy systems and cloud environments.
Integration, Automation, AI as ‘Superpowers’
The session “From APIs to AI Agents: The Future of Integration,” led by Diane Kesler, a Salesforce certified instructor, co-founder of Integration Quest, and MuleSoft practice lead at CloudBlazer, and Shivani Marrero, a senior MuleSoft developer, relayed this core message: Integration is shifting from API plumbing to AI-driven automation, and practitioner roles must shift with it.
“We all hear about integrations, automations, and AI independently,” Marrero said. But when teams “take all three together and create solutions, it’s not just improvement to your environment, it’s actually entirely new enterprise superpowers.” In that world, she argued, developers are no longer just integration specialists: “You’re going to be automation innovators.”
With that encouragement to evolve from “integration experts” to “automation innovators,” they mirrored Salesforce’s broader plan to merge integration, workflow automation, and AI services into one extensible configuration layer. That aligns with Salesforce’s published guidance, which highlights metadata as the context layer that informs agent behavior, APIs as the modular interfaces that allow workflows to be layered onto systems, and event streams as mechanisms for near-real-time orchestration. The session translated this structure into practical implications for skill development and tool fluency.
MuleSoft as ‘Superhighway,’ App Store for Data
They anchored that vision in MuleSoft’s core platform. “When I talk about foundation, it’s data, and when I talk about data, it’s MuleSoft,” Marrero said, adding that MuleSoft’s design center and API catalog are the mechanisms for turning system connectivity into reusable, productized assets. “MuleSoft is like a superhighway of your data.”
The API catalog, she added, should be “your company’s app store. You create the API…and anybody in your company can go view the description and utilize it. This is key because it creates reusability of APIs in your system.” System APIs then standardize how data is exposed, with MuleSoft flows transforming “any format” of data and “connecting to any system.”
That catalog does not stay confined. Once a trusted, bidirectional connection is established, the “API catalog is available in your Salesforce dashboard,” Marrero explained. Admins can choose which endpoints to surface, embedding integration into day-to-day Salesforce development rather than treating it as a separate discipline.
RPA as Part of the Automation Continuum
Marrero defined an AI agent as “an autonomous worker that understands your request in the natural language format, can think about your request, and execute it.” Topics define the scope and “different actions” available to an agent while “it’s using your reasoning engine. That’s how it’s becoming intelligent, creating critical tasks and thinking to get you answers.”
Kessler and Marrero then emphasized that RPA sits alongside APIs as an intentional extension point for automation in environments that still rely on screens, legacy sites, or file servers without modern interfaces. “AI agents are smart, but there are times when we have to interact with a…website or FTP server file to retrieve data from where an API doesn’t exist. Robotic process automation [RPA] allows us to create a bot,” Kessler said.
The session walked through three RPA examples: extracting rebate amounts from a legacy web page; downloading daily transport lists from a site; and automating clinical credential verification against the American Heart Association, Red Cross, and state licensing portals. In each case, bots followed “step by step instructions” to log in, navigate, extract data, and download artifacts.
“It’s not always present in any and every automation solution, but it’s very powerful,” Kessler noted, adding that once approved, bots run as secured sessions on Windows servers and “do all the work.”
In the automotive demo scenario, the bot’s output flowed back into Salesforce as structured data, reinforcing Kessler’s point that RPA is a designed component in the automation stack rather than a one-off patch.
A Unified Automation Stack
The speakers outlined how Salesforce expects customers to operationalize its Agentforce strategy by layering AI atop existing integration investments. MuleSoft’s API Manager, Composer, Anypoint Platform, and RPA were presented as a combined foundation for agent-enabled processes.
MuleSoft’s current guidance reinforces that direction, showing how existing Process and Experience APIs can be adapted into Model Context Protocol (MCP) servers to give agents structured, real time access to enterprise data and actions without additional architectural components. The approach also encourages teams to decompose complex endpoints into smaller tools that align with how agents sequence and execute tasks.
Taken together, that reflects MuleSoft’s wider vision of “universal connectivity,” where APIs expose system actions, automation tools assemble them, and AI determines when and how to trigger them. MuleSoft’s MCP connector, Flex Gateway controls, and Agent Fabric management layer add the governance and observability required for agents to interact with operational systems at scale. These developments position the integration layer as the execution substrate for AI-driven workflows, built on the same assets organizations already use for connectivity.
What This Means for ERP Insiders
Unified automation architectures are becoming strategic platforms. As Salesforce ties MuleSoft APIs, RPA, and Agentforce into one continuum, ERP providers and partners face rising expectations to expose clean, governed interfaces that external automation and agent layers can orchestrate. The highlighted bidirectional catalog integration and flow-based actions show the importance of API productization and event frameworks across core ERP suites.
Integration roles are shifting toward cross-disciplinary automation competencies. The reframing of integration practitioners as “automation innovators” reflects a broader convergence of integration engineering, workflow design, and AI operations. ERP product and services teams may need to adjust hiring profiles, training paths, and partner enablement models to match a world where developers must understand APIs, RPA, flows, and agent guardrails as a cohesive practice.
RPA gains a more formal role in AI-ready architectures. Validating RPA as an essential complement to APIs rather than a legacy patch signals a future of hybrid automation that spans modern and non-modern ERP estates. Program and platform leaders should assess how bots that run repeatable, well-governed tasks can complement API-led connectivity, especially where AI agents require reliable access to systems that lack API maturity but still sit in critical finance, HR, or supply chain workflows.
*Editor’s note: This article was updated on December 22 to include the session’s title and recording link.
