Salesforce announced more than 30 new AI capabilities for Slack on March 31, 2026 — the most sweeping overhaul the platform has seen since Salesforce acquired it for $27.7 billion in 2021. The strategic framing is direct: Slack is no longer a messaging tool. It is an agentic operating system.

The centrepiece of the update is a rebuilt Slackbot that now functions as a full enterprise AI agent. It does not just answer questions — it takes action, coordinates with other agents, and connects to data sources across a company’s entire stack.

What changed

The headline feature is AI Skills — reusable task templates that teams define once and deploy on demand. A skill captures the instructions, steps, and output format for a repeatable job: writing a budget summary, creating a project brief, drafting a customer follow-up. Once built, anyone on the team can trigger it in any context without re-explaining what needs to happen.

Alongside that, Slackbot is now a Model Context Protocol (MCP) client. This means it can connect to external services — Agentforce, third-party apps, enterprise data sources — and route work to specialised agents without a human driving the handoff. MCP is increasingly becoming the standard pipe for how AI agents talk to each other and to enterprise systems.

The update also includes:

• Live meeting intelligence — transcription and summarisation across any video provider, with Salesforce CRM records surfacing proactively when a customer name is mentioned
• Desktop monitoring — Slackbot operating outside the app itself, tracking deals, calendar, and conversations to proactively suggest actions
• Lightweight CRM for smaller teams — structured account data combined with Slack conversation history
• A third-party agent ecosystem — partners including OpenAI, Anthropic, Google, Perplexity, Notion, Dropbox, and Cursor connecting via a new RTS API and MCP server

The update builds on a January 2026 rollout that first gave Slackbot agentic capabilities. These 30 features represent the full acceleration of that direction.

The numbers behind the claim

The announcement came with internal data from Salesforce’s own deployment. The company has 86% of its employees using Agentforce in Slack. Their Engineering Agent handled 18,000 conversations across 3,500 users in six months, saving an estimated 30 hours per channel per month. Their Sales Agent is projected to save 203,000 hours annually.

Across the company, Salesforce reports 500,000 employee hours saved via Agentforce agents in Slack, with internal productivity value estimated at $6.4 million. Individual employees report saving between 2 and 20 hours per week.

These numbers will read differently depending on your instincts. Internal company statistics reported at launch events are worth treating with some scepticism. But the direction they point to is consistent with what teams deploying AI agents across other platforms are reporting. The variance is real — some use cases unlock dramatic time savings, others are still finding their footing.

The competitive frame

Salesforce CEO Marc Benioff has called Microsoft Copilot “Clippy 2.0” and “a disappointment” repeatedly in the past six months. That positioning is strategic. Copilot requires a $30 per user per month add-on for M365 customers. The upgraded Slackbot is included at no extra cost in Business+ ($15/user/month) and Enterprise+ plans.

The pitch is straightforward: you are already paying for Slack, and you now get an enterprise-grade AI agent at no additional cost, while a comparable Microsoft product costs double your current seat price and requires its own adoption curve.

Whether that framing holds up in practice depends heavily on what each team actually needs. But cost comparisons are an increasingly important part of how businesses evaluate AI tool spend, and Salesforce is playing that angle hard.

What This Means for Business

There are two things worth pulling out of this announcement for anyone thinking about their AI tooling.

First, the operating system question is real. Where does AI actually live in your organisation — in a standalone tool, inside your communication platform, embedded in your CRM? Salesforce is betting it lives in the collaboration layer. Microsoft has been betting the same thing. The platform where your team already spends time is where agentic capability has the most surface area, because agents need context and context lives in conversation.

Second, MCP is becoming infrastructure. The fact that Slackbot is now an MCP client, and that the partner ecosystem includes OpenAI, Anthropic, Google, and Cursor, signals that the protocol is cementing as the standard for how agents connect. If your team is building internal AI tooling, paying attention to MCP compatibility now will matter.

For businesses not yet on Slack or already locked into the Microsoft ecosystem, this announcement changes very little in the short term. For teams already using Slack who have been curious about agentic AI but not ready to build custom infrastructure, the updated Slackbot is the lowest-friction entry point they have had.

If this is the kind of problem agents can help with, the free Working With Claude field guide is the practical next step. Thirty-two pages, no fluff. Get the free guide.