Farewell to the “communication platform,” the “hub for your team and your work,” or the “replacement for email”: Slack now presents itself as an “agentic OS” (agentic OS).
The company is betting on this angle at Dreamforce 2025 (October 14–16). It touts a “transformation into a conversational workspace where people, AI, and agents collaborate, with the context of conversations and data.”
Dreamforce 2024: From Conversational to Agentic
A little over a year ago, at Dreamforce 2024 (September 17–19), the topic wasn’t yet an agentic OS, but a conversational work OS. Translated in alternative French renderings, the idea was the same as today: a “conversational interface that brings together teams, data, apps, and agents in a dedicated environment.”
At that moment, the Agentforce offering in Slack wasn’t yet available (it would enter beta in October 2024). The messaging focused, on one hand, on installing “agentic” apps via the marketplace (Claude and Perplexity were cited as examples, with a promise of imminent availability). On the other hand, on designing custom agents using the APIs.
Setting aside the agentic aspect, the Slack AI branding was generously promoted. The features gathered under this umbrella are now mapped into the paid Slack plans:
- Pro (€8.25 per user/month)
Channel and thread summaries, team call notes (capturing key information in a canvas), compatibility with third‑party AI assistants. - Business Plus (€18)
The same features plus daily AI recaps, file summaries, translation and explanation of messages, a natural-language workflows generator, writing assistance in canvases, and personalized search (based on conversations and shared files). - Enterprise+
The same features with, in addition, enterprise search (leveraging apps, databases, and systems connected to Slack).
The integration of these features was cited, mid-2025, to justify the price increase for the Business+ plan. Slack’s marketing messaging then hovered between the conversational work OS and the agentic OS: the aim was a “professional operating system for the AI‑agent era” (work operating system for the agentic era).
From General to Specific
With Agentforce in Slack reaching general availability in early 2025, it is now at the center of the messaging. The fairly generic description given at Dreamforce 2024 has given way to highlighting concrete use cases: for sales (Agentforce Sales), IT support (Agentforce IT Service), HR management (Agentforce HR Service), and data visualization (Agentforce Tableau).
Meanwhile, a general-use option is promoted through Channel Expert. This agent, activatable across all channels, can currently leverage conversations, canvases, lists, text files, and PDFs. It requires an Agentforce license.
Slack is also addressing developers. It announces the availability of an MCP server, initially granting access to conversations, files, and canvases. It also mentions a recently added element to its API: a block that facilitates displaying tabular data within messages.
With AI, an API That Is Less Open
The API has recently undergone a more notable evolution, a corollary of a data-access restriction approach. Slack indeed updated its terms of use in late May, primarily to prevent large-scale data exports. The company banned long-term storage and indexing, while noting that such data could not be used to train LLMs. In parallel, it introduced rate limits (requests per minute and messages per request) on the conversations.history and conversations.replies methods for non-marketplace, commercial apps.
The RTS (Real-Time Search) API, promoted from beta to “limited availability” at Dreamforce, fits this logic: it lets you leverage data without taking it out of Slack. Apps like ChatGPT, Google Agentspace, and Dropbox Dash use it. Perplexity Enterprise connects to the MCP server.
Originally tied to the Slack AI family, Slackbot is, on the surface, leaning toward agentic. In time, it should “perform actions on your behalf and build agents on your request,” we’re told. For now, one will turn to the Agent Builder component, potentially leveraging the few templates recently added (Customer Insights, Employee Help, Onboarding).
For additional context, read a note on Salesforce branding evolution, between chatbots, copilots, agents… and the dilution of the Einstein brand.
