Salesforce today generally made available an artificial intelligence (AI) agent for its Slack messaging platform that can organize work, create content and, soon, schedule meetings.
A revamped Slackbot now provides end users with a personal assistant that can be invoked via a conversational interface, says Rob Seaman, chief product officer for Slack at Salesforce. “We think Slack is the natural place for humans and agents to come together,” he says.
Based on the same Agentic AI framework that Salesforce is implementing across its software-as-a-service (SaaS) application portfolio, Slackbot now scans the interactions each end user has previously had on Slack to gain the context needed to organize activities, says Seaman.
Additionally, Slackbot takes advantage of existing integrations to perform tasks involving other Salesforce applications and will soon be able to take advantage of the Model Context Protocol (MCP) to access third-party data and interact with other AI agents, adds Seaman. The permissions assigned to that AI agent are inherited from the end user who creates the AI agent to ensure that it only performs the task assigned, he adds.
Finally, Salesforce also plans to make available a library of reusable prompts that end users can take advantage of to automate routine tasks. The overall goal is to provide end users with a personalized AI team mate that understands context in a way that enables it to continuously adapt to the end user, says Seaman.
That capability will make Slackbot more useful than a general-purpose AI agent that doesn’t understand how, for example, an end user prefers to interact with other employees, he notes.
Salesforce has decided to make Slackbot available as part of its existing licensing options rather than to charge extra for this capability. That approach should encourage more organizations to take advantage of Slackbot at a time when leading providers of enterprise applications are already jockeying for AI agent supremacy.
It’s not clear how many AI agents each end user might wind up employing, but Salesforce is betting that in organizations that have adopted Slack, this AI agent will be relied on daily to the point where they become indispensable. Of course, Salesforce is not the only organization with similar ambitions so each organization will need to decide which mix of AI agents to deploy, many of which are likely to have overlapping capabilities.
Regardless of approach, AI agents will soon be pervasively employed across the enterprise, which, in addition to improving productivity, also serves to increase the size of the attack surface that needs to be defended. Each agent is, in effect, a new type of end user that has permissions to access a wide range of backend end services. IT teams will need to keep track and manage AI agents in much the same way they would a human employee.
Naturally, each end user will attain varying levels of proficiency when it comes to making use of AI agents, and many organizations would be well advised to provide some level of training rather than simply unleashing them without fully understanding the potential opportunity and inherent risks.
