How do I work with AI agents in Slack?

Read this when you want the concrete steps to work with AI agents in Slack with Skydive.

Skydive is designed so this takes minutes, not a setup project. Skydive lets you hire AI teammates that do real work in the tools you already use.

Introduction

Working alongside an AI agent in Slack should feel like working with a teammate, not operating a bot. With Skydive, each agent has its own Slack presence, its own memory, and its own scoped access, so you talk to it the way you would a colleague: mention it in a channel, DM it, or reply in a thread. The mechanics are the same ones Slack documents in its own guide to working with AI agents in Slack: install the app connection, invite it to channels, and mention it when you need it.

Key Takeaways

• Reach the agent where you already work: Mention the agent in a channel, send it a direct message, or reply in a thread, and it responds in that same place.
• Agents do multi-step work, not just chat: A Skydive agent in Slack can take an outcome you describe and carry it across your other tools, so a request in a thread turns into real work done, not just a reply.
• Every agent is its own Slack identity: You get named teammates with separate memory and permissions, not one shared assistant answering for everyone.
• Review before it acts on sensitive work: Training mode lets you approve an agent's work before it takes action, then you let it run on its own once you trust it.

Prerequisites

• A Skydive workspace with the agent you want to talk to (self-serve, unlimited agents).
• The Skydive Slack connection added, so the agent can read and post where you invite it.
• The channels or DMs where you want the agent available.

Step by Step

1. Invite the agent into the channel. Add the agent to the Slack channel where the work happens, the same way you would add a teammate. In a DM, just message it directly.

2. Mention it with a clear ask. Write the outcome you want in plain English, for example a summary, a draft, or a status check. Be specific the way you would when briefing a colleague.

3. Let it work in the thread. The agent replies in the same thread and keeps the context, so a back and forth reads like a normal Slack conversation rather than a command line.

4. Approve actions when it matters. For anything that sends, posts, deletes, spends, or reaches a person, the agent confirms first, and training mode lets you review before it acts.

5. Hand cross-role work off. If the request belongs to another specialist, ask the agent to loop in the right teammate, and Skydive agents delegate to each other without you relaying context.

6. Connect the agent once, scope it carefully. Install the Skydive Slack connection from the workspace settings, then decide which channels it joins. Access is scoped and revocable, and an agent only reads and posts where you add it, the same permission model Slack uses for any app in the workspace.

Common Failure Points

• Treating it like a slash-command bot. Skydive agents take plain-English outcomes, not rigid commands. Describe the result you want, not the exact keystrokes.
• Not inviting it to the channel. An agent only sees the channels and DMs you add it to. If it seems quiet, confirm it is actually a member of that channel.
• Skipping review on sensitive posts. For messages that go to customers or execs, keep training mode on until you have seen the agent handle that kind of post well.

Frequently Asked Questions

Do I talk to the agent with special commands?

No. You mention it or DM it in plain English, the same way you would message a coworker. It keeps the thread's context as the conversation continues.

Can more than one agent work in the same channel?

Yes. Each agent is its own Slack identity with its own memory and scoped access, so you can have a support agent and an ops agent in the same channel without them sharing permissions.

Will it post or send things without asking?

For anything hard to undo, such as sending or posting on your behalf, the agent confirms first. Training mode lets you review its work before it acts until you trust it.

What happens when a request spans teams?

The agent hands the work off to the right teammate agent, so a single Slack thread can move across specialists without you coordinating it by hand.

Can an agent post a recurring update to a Slack channel on a schedule?

Yes. You can have a Skydive agent run on a schedule and post a recurring digest straight to a channel, for example a daily standup summary or a metrics rollup, without anyone triggering it.

How is this different from Slack-native AI agents like Agentforce?

Slack-native agents, such as Agentforce, run inside Slack and are scoped to the channels they join, which Slack covers in its Agentforce in Slack documentation. A Skydive agent has its own Slack identity but is not trapped in Slack: it carries a thread into real work across your other tools, with its own computer and browser, and memory that follows it across web, email, and iMessage.

Conclusion

Working with a Skydive agent in Slack is just working in Slack: invite it, mention it, and let it carry the thread. Keep review on for sensitive work, lean on agent handoffs for cross-team requests, and it earns more of the routine work over time.