Salesforce today revealed it has made generally available a platform that makes use of artificial intelligence (AI) to re-engineer and automate back office workflows.
Launched at a Salesforce Agentforce World Tour event in New York, the Agentforce Operations platform is based on a set of specialized AI agents that have been trained to autonomously complete tasks such as extracting data from complex documents, running computations, updating credit scores, and identifying compliance gaps.
Based on a platform that Salesforce gained with the acquisition of Regello late last year, Agentforce Operations can be used to either create a new workflow or modernize an existing one based on, for example, an email workflow, says Sanjna Parulekar, director of product marketing for Salesforce.
For example, Agentforce Operations can turn unstructured documents or diagrams into a set of auditable digital blueprints in a few minutes that AI agents will execute. Alternatively, organizations can make use of more than 30 out-of-the-box blueprints for common tasks such as invoice auditing and new customer onboarding. Those AI agents can also proactively flag delays, such as a three-day lag in an approval, and even suggest work arounds.
Salesforce previewed Agentforce Operations at its recent Dreamforce event. Unlike other elements of the Agentforce portfolio that tend to focus more on customer experiences, this latest addition is squarely focused on modernizing back office workflows that today are generally highly fragmented, says Parulekar.
In contrast, the AI agents from Salesforce are designed to not only complete a task, but also coordinate with people only when needed. It will be up to each organization to determine when humans will need to be in a workflow loop to check on whether those tasks have been completed, notes Parulekar.
Additionally, Salesforce provides a set of observability capabilities designed to make it possible to audit how AI agents have completed any given set of tasks, she adds.
The overall goal is to make it simpler to employ AI agents in a way that ultimately serves to make organizations more efficient, says Parulekar.
Salesforce is now encroaching on territory now held by ServiceNow, SAP, and Oracle’s territory,” says Keith Kirkpatrick, vice president and research director for enterprise software and digital workflows for the Futurum Group. If done well, this approach should drive more business value by focusing on complete back-end workflows in a way that delivers a higher return on investment for the Agentforce platform, he adds.
In general, Salesforce is betting that its ability to govern, manage and orchestrate AI agents will spur further adoption of its platforms. The company claims to already have driven consumption of more than 19 trillion AI tokens and has developed an Agentic Work Unit (AWU) to enable organizations to better track returns on investment (ROI).
That latter unit is critical because while Salesforce enables employees to access AI agents for a flat fee, it charges for any external access. Therefore, IT teams that are deploying third-party AI agents and Model Context Protocol (MCP) servers that invoke headless Salesforce services will need to factor those costs into their overall strategy. Ultimately, however, Salesforce is betting most organizations will opt to consume the AI agents that it develops that by definition will be more deeply integrated into its data model. “A lot of customers have quickly abandoned do-it-yourself efforts using third-party MCP servers in favor of using our agents,” says Parulekar.
Regardless of approach, the one thing that is certain is there will soon be armies of AI agents that are accessing a wide range of headless services. The challenge, of course, is that depending on who provides them, not all those AI agents will necessarily have the same level of access and, just as importantly, level of control.
