Slack Just Became an AI Work Assistant With 30 New Features

Salesforce just dropped the biggest Slack update in years — 30 new AI features that transform the messaging app from a place where you talk about work into a platform where work actually gets done autonomously. The new Slackbot, powered by Salesforce's Agentforce platform, can automate workflows, manage CRM data, summarize meetings, generate reports, and proactively surface information you need before you even search for it.

This is not a chatbot that answers questions. This is an autonomous work agent that lives inside the communication tool your entire team already uses daily. The distinction matters because it means AI assistance is no longer a separate app you context-switch to — it is embedded in the flow of work. For the 35+ million daily Slack users across enterprise teams, that shift is enormous.

---

The 10 Features That Change Everything

1. Automatic Meeting Summaries: After every Huddle or connected Zoom call, Slack generates a comprehensive summary with key decisions, action items, and follow-ups — automatically, without anyone asking. The summaries are surprisingly good, capturing nuance and attribution correctly. Action items are auto-assigned to channel members with due dates.

2. Thread Compression: Long discussion threads that span 50+ messages get compressed into key decisions and action items. Instead of scrolling through an hour of back-and-forth, you get the three things that actually matter in a two-paragraph summary. The "tldr" command works on any channel or thread.

3. Proactive Intelligence: Slack now monitors your channels and surfaces relevant information before you search for it. Upcoming deadlines mentioned in conversations, customer issues escalating in support channels, team blockers that are slowing projects — they appear as gentle notifications at the right time, never noisy.

4. Workflow AI Steps: The workflow builder now includes AI decision nodes. You can create automated sequences where AI drafts emails based on trigger conditions, analyzes incoming data, and makes routing decisions within automated pipelines. This turns Slack into a legitimate process automation platform — without writing a line of code.

5. CRM Integration: For Salesforce users, the new Slackbot can pull customer data, update records, log activities, and generate account summaries directly from chat. Ask "what is the status of the Acme deal?" and get a comprehensive CRM summary without leaving Slack.

6. Semantic Channel Search: Search now understands intent, not just keywords. Asking "did anyone mention the Q3 launch slipping?" returns relevant messages even when the wording is different. It scans across channels you have access to and ranks by relevance, not recency.

7. AI Drafted Replies: When you open a long thread, Slack offers a draft response that captures your tone based on past messages. You can accept, edit, or discard. Power users save 30+ minutes a day on reply triage alone.

8. Document Q&A: Drop a PDF, spreadsheet, or Google Doc into any channel and ask questions in plain English. Slack reads the file, indexes it, and answers from the content. This is similar to how Claude integrates directly into Excel — but conversational and shared across the team.

9. Channel Recaps on Demand: Coming back from time off? Run "/recap #channel since:monday" and get a structured digest of decisions, blockers, and open threads. The recap respects access permissions and only summarizes what you can already see.

10. Open Agent Connectors: Slack now supports MCP (Model Context Protocol) connectors, letting external AI agents talk to Slack channels with proper scopes and audit logs. If you have ever wondered why MCP matters, this is the use case — read our MCP explainer for context.

Real Workflow Examples

Summarizing a 200-message thread before a standup. A product manager runs "/summarize" on the overnight engineering thread. In 3 seconds she gets: the deploy that broke at 2am, the rollback that fixed it, the open question about feature flags, and the two engineers who own follow-up. What used to be a 20-minute scroll-and-skim is now a glance.

Auto-generated meeting notes that route themselves. A sales call ends in Huddle. Slack writes the summary, tags the deal in Salesforce, creates three follow-up tasks (proposal draft, legal review, demo scheduling), and pings the people who own them — all before the AE has closed their laptop.

Cross-channel search that actually finds things. An engineer searches "the database migration plan we discussed last quarter." Old Slack returned 400 results sorted by date. New Slack returns the three messages that contain the actual plan, plus a generated summary of what was decided and what got punted.

Slack AI vs Microsoft Teams Copilot

Microsoft Teams Copilot has been live for 18 months and remains the enterprise default for organizations on Microsoft 365. It excels at summarizing meetings, drafting in Word, and pulling from SharePoint. But Copilot is a layer bolted onto Office; Slack's new agents are native to the channel model and connect deeply to Salesforce CRM data — something Teams cannot match without a third-party plug-in.

Teams Copilot costs $30/user/month on top of M365. Slack AI is included in Business+ and Enterprise Grid, with the Agentforce agents priced per conversation. For sales-led organizations already on Salesforce, Slack now becomes the obvious choice. For engineering or Microsoft-shop companies, Teams Copilot still wins on document integration.

Privacy and Data Considerations

Salesforce confirmed that Slack AI does not train foundation models on your messages. Customer data stays inside the Einstein Trust Layer, which masks PII before any LLM call and audits every prompt and response. Admins can scope AI access per channel, exclude private DMs entirely, and require legal hold rules to apply to AI outputs the same as any other message.

That said, enabling Agentforce means giving an autonomous agent permissions to act in CRM. Audit logs are essential, and most security teams will want to start with read-only scopes before allowing writes.

Pricing Tier Guide

Pro ($8.75/user/month): Basic AI summaries and recap. No workflow AI steps, no Agentforce.
Business+ ($15/user/month): Full Slack AI — summaries, search, drafted replies, document Q&A.
Enterprise Grid (custom): Adds Agentforce agents, MCP connectors, advanced governance, and unlimited AI workflow steps.
Agentforce add-on: $2 per agent conversation, billed monthly. Volume pricing for enterprise.

FAQ

Is Slack AI free?
No. Slack AI is included on Business+ and Enterprise Grid plans. Pro users get a limited subset (basic recaps only). Agentforce agents are a paid add-on billed per conversation.

Does Slack train its AI models on my company's messages?
No. Salesforce states that customer data is not used to train foundation models. PII is masked at the Einstein Trust Layer and every AI call is logged for audit.

Can Slack AI replace a Microsoft Teams Copilot deployment?
For Salesforce-centric or sales-led teams, yes — the CRM integration is the killer feature Teams cannot match. For Microsoft-shop engineering organizations, Teams Copilot remains stronger on document workflows.

What is Agentforce and how is it different from Slackbot?
Slackbot is the assistant you chat with. Agentforce is the platform behind it — a set of autonomous agents with scoped permissions that can act in Salesforce, run workflows, and call external systems via MCP. Think of Slackbot as the face, Agentforce as the brain.