Group Chat Agent for Slack

The Group Chat Agent lets employees get answers from your AI Colleague directly inside Slack channels, private channels, and group DMs — without leaving the conversation. Anyone in the channel can @-mention the AI Colleague, and it responds with full awareness of the ongoing discussion.

Previously, the AI Colleague on Slack only worked in 1:1 direct messages. With the Group Chat Agent, the AI Colleague becomes a true participant in team conversations — it can be tagged in a channel or a thread, reads the recent discussion for context, and delivers the answer either publicly in the channel or privately in the employee's DM, depending on the nature of the answer.


How It Works

1. Mention the AI Colleague

Any member of a channel where the AI Colleague has been added can tag it with a question:

@Leena AI HR Assistant what's the leave carry-forward policy for this year?

Mentions work in three places:

SurfaceBehavior
Channel (root message)The AI Colleague replies in a thread under the mention
Inside a threadThe AI Colleague replies within the same thread
Group DM (multi-person DM)The AI Colleague replies in the group DM

2. Instant acknowledgment

As soon as the AI Colleague is tagged, it posts a placeholder message ("Working on it...") in the channel or thread, so everyone knows the request has been picked up.

3. Context-aware processing

The AI Colleague reads the recent conversation around the mention so it can understand references like "can you summarize the issue above?" or "what's the deadline they mentioned?":

  • Channel mentions: Up to the last 30 messages from the past 6 hours in the channel
  • Thread mentions: The full thread the mention was posted in
  • System messages (joins, leaves, topic changes, pins) and deleted messages are automatically excluded
  • User IDs in the context are resolved to display names so responses read naturally

4. Answer delivery — public or private

The AI Colleague intelligently decides where the answer belongs:

Delivery modeWhat happensWhen it's used
Public (in-channel)The placeholder is updated with the full answer, including knowledge sources and an AI disclaimerThe answer is general and appropriate for everyone in the channel
Redirect to DMThe placeholder is updated with a short notice (e.g., "Got it, Priya — the answer here is specific to you, so I've shared it in your DM."), and the full answer is delivered in the employee's DMThe answer contains personal/private content, or resolving the query needs a multi-step back-and-forth

In both cases, the employee who asked also receives the full answer in their direct message with the AI Colleague, with a link back to the original channel thread ("Here's the answer to your question in #channel-name · view thread").

5. Follow-ups continue in DM

The employee can continue the conversation in their DM thread — the AI Colleague retains the context of the original channel question, so follow-ups like "and what about contractors?" work seamlessly.

6. Attachments

If the @-mention includes file attachments (documents, images), the AI Colleague processes them as part of the question. Files shared in the channel without tagging the AI Colleague are ignored — the agent never responds to messages that aren't addressed to it.


Enabling the Group Chat Agent

Prerequisites

  1. Slack channel integration is configured for your AI Colleague (App ID, Signing Secret, and Bot User OAuth Token installed). See Slack Setup for the base integration.
  2. Generative AI (GPT) responses are enabled for the AI Colleague.
  3. Response streaming is enabled for the AI Colleague.
  4. The Slack app's OAuth scopes and event subscriptions include the permissions listed below (re-install the app to the workspace after adding new scopes).

Note: If GPT or streaming is not enabled, @-mentions in channels are not processed by the Group Chat Agent and fall back to standard behavior.

Adding the AI Colleague to a channel

The AI Colleague must be a member of the channel or group DM to respond there:

  1. In the Slack channel, type /invite @<Your AI Colleague name>, or
  2. Open channel settings → IntegrationsAdd apps → select the AI Colleague app

Once added, any channel member can @-mention it.


Slack Permissions Required

Bot token scopes

Configure these under OAuth & Permissions in your Slack app (or use the auto-generated Leena AI app manifest, which includes the base scopes):

ScopeWhy it's needed
app_mentions:readReceive @-mentions of the AI Colleague in channels and threads
chat:writePost the placeholder, the answer, and the redirect notice
channels:historyRead recent messages in public channels for conversation context
groups:historyRead recent messages in private channels for conversation context
mpim:historyRead recent messages in group DMs for conversation context
channels:readResolve public channel names for references and DM links
groups:readResolve private channel names for references and DM links
im:history, im:writeDeliver and continue the answer in the employee's DM
users:read, users.profile:read, users:read.emailResolve user IDs to display names and map Slack users to employee profiles
files:read, files:writeProcess attachments shared with the mention
links:readUnfurl Leena AI links in responses
assistant:writePower the Slack AI Assistant experience (status, suggested prompts)

Event subscriptions

Under Event Subscriptions → Subscribe to bot events, ensure the following are present (the request URL is the standard Leena AI Slack webhook):

EventPurpose
app_mentionTriggers the Group Chat Agent when the AI Colleague is tagged
message.imPowers DM answers and follow-up continuity
link_shared, app_uninstalledStandard integration events

Important: After adding scopes or events, click Reinstall App in Slack so the new permissions take effect. Workspace admin approval may be required depending on your Slack workspace settings.


Privacy & Data Handling

  • Personalized answers stay private. If the answer depends on the employee's own data (e.g., their leave balance, payslip), it is never posted in the channel — the channel only sees a redirect notice, and the answer goes to the employee's DM.
  • Only addressed messages are processed. The AI Colleague responds exclusively to messages where it is explicitly @-mentioned. Regular channel chatter and unaddressed file uploads are never processed or answered.
  • Bounded context. Channel context is limited to a recent time window and message/token caps; it is used only to answer the specific question asked.
  • Access control applies. Knowledge base answers respect the asking employee's permissions — the same access rules as in 1:1 DM conversations.

Limitations

  • Slack Connect / shared channels are not supported. Mentions in channels shared with external organizations are ignored by design.
  • The Group Chat Agent requires GPT + streaming to be enabled; classic (non-generative) workflow bots do not respond to channel mentions via this flow.
  • The AI Colleague must be a member of the channel to receive mentions and read context.
  • Channel context is capped (last 6 hours / 30 messages for channel mentions); very old messages are not considered.

Troubleshooting

SymptomLikely causeFix
AI Colleague doesn't respond to @-mentions in a channelIt's not a member of the channel/invite the app to the channel
Placeholder appears but is replaced with "Sorry — I could not complete that request."The AI engine couldn't process the requestRetry; if persistent, check the AI Colleague's GPT configuration
No response at all in channels, but DMs workGPT or streaming is disabled, or app_mention event / app_mentions:read scope is missingEnable GPT + streaming; verify scopes and events, then reinstall the app
Answer lacks channel context or channel name shows as a raw IDHistory/read scopes (channels:history, channels:read, etc.) missingAdd the scopes and reinstall the app
Mentions in a Slack Connect channel are ignoredShared channels are unsupportedUse an internal (non-shared) channel