Slack’s AI updates signal shift towards agent orchestration

“Slack is increasingly being positioned as the interface where AI agents coordinate tasks across systems, workflows and teams,” she said, indicating a broader shift in how organizations are approaching AI. “The focus is moving away from individual tools and toward systems that connect processes, data and applications.” Slack announced more than 30 features on Tuesday, many of which are aimed at making its Slackbot AI agent more effective and easier to access. This means advancing Slackbot from a “simple helper that helps you find information and synthesize information” to a “teammate,” said Rob Seaman, Slack’s general manager, in a news briefing. Uses can now talk to Slackbot rather than typing, for example, while a memory function enables it to remember user preferences and interactions across sessions. Slackbot can now search the web for information too, rather than just searching within a Slack workspace and connected apps. The AI agent can join video and voice meetings to take notes and interact with meeting participants. Slackbot can take actions during a call, such as presenting a Salesforce record when a client’s name is mentioned or updating the CRM record as directed, the company said. A desktop integration enables Slackbot to access what a user is working on in a different app. When directed, Slackbot takes a screenshot for context to carry out actions set out by the user, such as viewing an email in a separate app and getting Slackbot to draft a response. Coordinating layer It’s the addition of a model context protocol (MCP) client option that points to a potential future for Slack as a coordinating layer for getting work done, though. This will enable Slackbot to interact with Agentforce and external agents more easily. “Now it can route work or prompt questions to any agent or app that’s in your Slack,” determining which agent is most appropriate for a particular task, said Seaman. The launch of the MCP client points to Slack becoming a “routing layer that decides which system or agent performs a task,” simplifying the user experience, said CCS Insight’s Bell. At the same time, it introduces another layer that must be “trusted, governed and integrated into existing controls,” and without these guardrails, there’s a risk that “complexity is redistributed rather than reduced,” she said. It can also require a shift in management approach. “For IT teams, this creates a new control point, where the priority is not the tools themselves, but how actions are authorized, tracked and, if needed, reversed across systems,” she said. Bell also pointed to a gap between AI functionality that software vendors develop and launch, and the ability of organizations to actually adopt it. CCS Insights research suggests that most enterprises are not yet structured to support the level of coordination afforded by more advanced AI tools. “Only a small minority have fully integrated automation strategies, and many initiatives remain at the pilot stage because workflows, data and governance are not mature enough to support scale,” said Bell. There’s often a gap between ambition and execution, she said, with senior leaders increasing investment in AI while their priorities remain anchored in tackling security, governance and operational stability. That means the success of that Slack’s attempts to move up the stack from communication layer to control and orchestration will depend less on the number of features announced and “more on how well Slack can operate within existing enterprise constraints such as identity, permissions, auditability and integration with legacy systems.” Slackbot is available on Slack’s Business+ and Enterprise+ paid subscription plans. For Business+ customers, individual users can send 15 messages to Slackbot each week, a spokesperson said, while Enterprise+ customers have unlimited usage.