Agent Slack Integration

Deploy your conversational AI agents directly in Slack channels for intelligent team collaboration

Overview

Agent Slack Integration lets conversational agents work in Slack channels. Messages are handled in Slack, and conversation data is tracked in your Elementum app where you configure permissions and workflows.

Slack is configured per channel, with a different agent per channel if you want. The Microsoft Teams integration instead uses one-on-one conversations and a single agent configuration.

---

Prerequisites

Organization admin access is required for the initial Slack app installation. Ensure you have proper permissions before starting.

You also need:

• Active agent: At least one agent in your app, with training appropriate for the channel
• Slack access: Ability to install or request apps in your workspace (see your workspace’s app policies)
• Channel access: Permission to add apps to the channels where the agent should run

---

Setting Up Slack Integration

1. Install Elementum App from Slack Marketplace

   Begin by adding the Elementum AI app to your Slack workspace through the official Slack App Directory. If you need help with Slack app installation permissions or approval flow, see Add apps to your Slack workspace.

   Installation Process:

   1. Click Add to Slack to install the Elementum app
   2. Review the permissions the app is requesting
   3. Click Allow to proceed with the installation

   The Elementum app requires standard permissions to read messages, post responses, and manage threaded conversations. These permissions are necessary for the agent to function properly within your Slack channels.

2. Complete Setup in Elementum

   Required Permission:

   After you install the app in Slack, you are redirected to Elementum to link the workspace to an organization.

   Organization selection

   1. On the Elementum setup page, select the organization to associate with this Slack installation.
   2. Click Complete Installation.
   3. Wait for the Installation Successful confirmation.

   The organization you select determines which app data and agents are available when you configure channels.

   To confirm the installation: in Slack, open Apps and verify Elementum Intelligence is listed. In Elementum, verify the integration appears as installed for the correct organization.

3. Add Elementum to Your Channel

   After installing the Elementum app, add it to the specific channel where you want the agent to operate.

   How to Add:

   Option A: Using /invite command

   1. Type: /invite @Elementum Intelligence
   2. Press Enter to add the bot to the channel

   Option B: Using @ mention

   1. Simply mention the Elementum bot (@Elementum Intelligence) in a message
   2. The app will automatically be added to the channel

   Option C: Through channel settings

   1. Click the channel name at the top of the channel
   2. Click the More icon, then select Open channel details
   3. Select Integrations → Add an app
   4. Search for and select the Elementum app you installed

   Wait for confirmation that the Elementum bot has joined the channel before proceeding.

4. Open setup in the channel

   In the channel where the agent should run, type /elementum setup and press Enter. The configuration dialog opens. The following steps mirror the sections in the dialog: Agent Selection, Conversation Settings, then Channel Messages.

5. Agent Selection

   Select an app to configure for this channel:

   • Choose the Elementum app that contains the agent you want to use
   • This determines which app environment the agent operates within

   Choose an agent

   • Select the specific agent that will respond to messages in this channel
   • Different channels can use different agents based on their purpose

   Continue in the dialog to Conversation Settings when you are ready.

6. Conversation Settings

   Select conversation timeout:

   • Choose how long conversations remain active before automatically expiring
   • Options range from minutes to hours depending on your needs
   • Helps manage agent workflows and conversation context

   Include timeout reminder messages:

   • Enable to send users warnings before conversations expire
   • Provides better user experience by setting clear expectations

   Interaction Options:

   • Quick Answers: Brief, direct responses to questions without creating app records.
   • Full Conversations: Multi-turn conversations with full context. Creates an app record to store the conversation and maintain workflow state across interactions.

   Continue in the dialog to Channel Messages when you are ready.

7. Channel Messages

   Channel message handling controls how the agent responds to messages:

   Interact with all messages

   Behavior: Agent responds to every message posted in the channel

   Best for:

   • Dedicated agent support channels
   • Low-traffic channels where agent participation is expected
   • Channels specifically created for agent interactions

   Considerations: May be overwhelming in high-traffic channels

   Interact with messages detected as questions

   Behavior: Agent only responds to messages containing a ’?’ question mark

   Examples:

   • Responds: “What is our current inventory status?”
   • Does not respond: “Please update the inventory report”

   Best for:

   • General team channels where selective agent participation is desired
   • Channels with mixed conversation types
   • On-demand agent assistance

   Considerations: Users must explicitly ask questions for agent engagement

   Interact with questions where there is a high level confidence answer

   Behavior: Agent only responds when it has high confidence in providing accurate answers

   Best for:

   • Channels requiring high accuracy standards
   • Reducing agent noise in busy channels
   • Ensuring quality over quantity of responses

   Considerations: Some valid questions may not receive responses if confidence is below threshold

   Do not interact

   Behavior: Agent is present in the channel but does not respond to any messages

   Best for:

   • Temporarily disabling the agent without removing it
   • Testing other configuration options
   • Maintaining agent presence while troubleshooting

   Note: Alternatively, you can remove the bot from the channel entirely with /remove @Elementum Intelligence

8. Complete configuration

   1. Review all settings in the dialog (Agent Selection, Conversation Settings, and Channel Messages).
   2. Click Next to finalize the agent setup
   3. Wait for confirmation that the agent has been successfully configured

---

Using Slack Integration

Channel-Based Agent Deployment

Each channel has its own agent, timeout, and Channel messages behavior. Match agent training to the channel’s purpose (for example, support vs. general questions). Example combinations appear in Managing multiple channel configurations.

Starting Conversations

For end users:

1. Open the channel where the Elementum agent is present.
2. Post according to the channel’s Channel messages setting (any message, questions with ?, or other rules you configured).
3. The agent replies in the channel or a thread; context lasts for the active conversation.

In Full Conversations mode, the user’s Slack identity is associated with the conversation record in your app.

Conversation Management

Quick Answers and Full Conversations are set in /elementum setup in the Conversation Settings step (see Setting up Slack integration). In short:

• Quick Answers: Short replies, no app record, no multi-turn state between questions.
• Full Conversations: App record per conversation, full transcript and context, suitable for automations and follow-up.

Timeouts close idle conversations after the time you set; optional reminders warn users first. After a timeout, the next message starts a new conversation (new record in Full Conversations mode).

---

Configuration Best Practices

Matching Settings to Channel Purpose

Dedicated Support Channels

Recommended Configuration:

• Interaction Mode: Full Conversations
• Channel Messages: Interact with all messages
• Timeout: 30-60 minutes for extended support interactions

Rationale: Users expect comprehensive help and continuous engagement in dedicated support channels.

General Team Channels

Recommended Configuration:

• Interaction Mode: Quick Answers
• Channel Messages: Interact with questions only
• Timeout: 15 minutes

Rationale: Minimizes agent noise while providing on-demand assistance in busy channels.

High-Stakes Channels (Legal, Compliance, Executive)

Recommended Configuration:

• Interaction Mode: Full Conversations
• Channel Messages: High confidence answers only
• Timeout: 45 minutes

Rationale: Ensures accuracy and reduces risk of misinformation in sensitive contexts.

FAQ/Knowledge Base Channels

Recommended Configuration:

• Interaction Mode: Quick Answers
• Channel Messages: High confidence answers only
• Timeout: 10 minutes

Rationale: Provides quick, accurate answers to common questions without extended conversations.

Agent Training for Slack

Train agents for public or semi-public channels: shorter replies, Slack formatting when it helps, threaded replies for long exchanges, and clear guidance to move sensitive topics to DM or email when appropriate.

Example Training Instructions

You are operating in a Slack channel with multiple participants. Keep responses
concise and professional. For sensitive information, suggest the user contact
you via direct message or email. Always acknowledge other participants'
contributions when relevant.

Conversation Timeout Strategy

Timeout Duration Recommendations:

Quick FAQs

10-15 minutes

Simple question/answer interactions

Standard Support

30-45 minutes

Typical support conversations with multiple exchanges

Complex Issues

60+ minutes

Technical troubleshooting or detailed consultations

In busy channels, prefer shorter timeouts and stricter Channel messages settings so conversations do not pile up.

---

Integration with Workflows

Automatic Record Creation (Full Conversations Mode)

In Full Conversations mode, starting a conversation creates an app record that typically holds Slack user details, start time, assigned agent, and channel or thread context. Use that record for review and downstream automations.

Workflow Triggers

Use the Agent Conversation Ended trigger in your automations to process Slack conversation outcomes. The trigger exposes the transcript, user details, timing and counts, messages, and channel metadata.

Automation Examples:

• Create follow-up tasks based on unresolved issues
• Send summary notifications to relevant team members
• Update CRM or support systems with interaction details
• Generate analytics reports on agent performance
• Route urgent matters to appropriate departments
• Trigger approval workflows for agent recommendations

Viewing Slack Conversations

After a Slack conversation ends (in Full Conversations mode), you can view the complete interaction:

Conversation Interface:

1. Navigate to the automatically created record in your app
2. Access the Chat/Conversation interface in record details
3. View the conversation marked as a Slack conversation type
4. Review complete transcript with timestamps and participant details
5. See channel context and threading information

---

Managing Multiple Channel Configurations

Channel-Specific Agent Strategy

Pilot in low-traffic or internal channels before wider rollout.

Strategic Agent Deployment:

• Specialized Agents: Deploy domain-specific agents to relevant channels
• Consistent Agents: Use the same agent across related channels for consistency
• Hybrid Approach: Mix specialized and general-purpose agents based on needs

Example Multi-Channel Setup:

Channel	Agent	Mode	Interaction
#customer-support	Support Agent	Full Conversations	All messages
#product-questions	Product Expert	Quick Answers	Questions only
#general	General Assistant	Quick Answers	High confidence only
#engineering	Technical Agent	Full Conversations	Questions only
#sales	Sales Assistant	Full Conversations	All messages

Reconfiguring Channels

To modify agent configuration for a channel:

1. Navigate to the channel where the agent is configured
2. Type /elementum setup to open the configuration dialog
3. Update any settings (agent selection, timeout, interaction mode)
4. Click Next to save the new configuration
5. The agent immediately adopts the new settings

Configuration changes take effect immediately without needing to remove or re-add the bot to the channel.

Monitoring Channel Performance

Metrics to Track

• Response Rate: How often the agent responds to messages
• Conversation Duration: Average length of agent conversations
• User Satisfaction: Feedback on agent helpfulness
• Channel Activity: Volume of messages and agent interactions
• Timeout Frequency: How often conversations expire before natural completion

Optimization Based on Metrics

• Adjust interaction mode if response rate is too high or too low
• Modify timeouts based on average conversation duration
• Refine agent training based on user feedback
• Change confidence thresholds based on response quality

---

Troubleshooting

Common Issues

Bot not responding in channel

• Verify the agent is properly configured using /elementum setup
• Check interaction mode settings (agent may be set to Do not interact)
• For Questions only mode, ensure messages include ? question marks
• Confirm the agent is active and functioning in your Elementum app
• Verify the bot has been added to the channel successfully

Agent responding to wrong messages

• Review the Channel messages setting and adjust interaction mode
• Consider switching from All messages to Questions only or High confidence only
• Refine agent training to better identify relevant questions
• Check if multiple agents are configured in the same channel (may cause conflicts)

Conversation records not creating

• Verify you’re using Full Conversations mode (Quick Answers doesn’t create records)
• Check that the target app is active and accessible
• Ensure proper permissions for record creation
• Review field mappings and data capture configuration

Timeout issues

• If conversations expire too quickly, increase timeout duration
• If timeouts are too long, reduce duration to better match actual usage
• Enable timeout reminder messages to improve user experience
• Monitor actual conversation patterns to optimize timeout settings

Configuration not saving

• Ensure you have proper permissions to configure the channel
• Verify the selected app and agent are valid and active
• Check that all required fields are completed
• Try removing and re-adding the bot to the channel if issues persist

Performance Optimization

High-traffic channels

• Use Questions only or High confidence only interaction modes
• Implement shorter conversation timeouts (10-15 minutes)
• Consider Quick Answers mode to reduce overhead
• Monitor agent load and adjust configuration as needed

Low-response channels

• Switch to All messages or Questions only mode to increase engagement
• Reduce confidence threshold if using high confidence mode
• Review agent training to improve question detection
• Adjust timeout to match actual interaction patterns

Agent training for Slack channels

See Agent Training for Slack. Tune prompts for channel noise, when to stay silent, and when to escalate to a person.

---

Security and Compliance

Data Protection

• User Information: Slack user data is handled according to your data retention policies
• Conversation Storage: All interactions are stored securely within your Elementum environment
• Access Control: Conversations are only accessible to authorized users within your organization
• Channel Privacy: Agents respect Slack channel privacy settings (public vs private channels)

Slack Security

• Organization Control: Aligns with who can install apps in your workspace (see Prerequisites)
• User Permissions: Follows your organization’s Slack app installation policies
• Data Residency: Conversations remain within your configured data boundaries
• Channel Permissions: Agents only access channels they’re explicitly added to

Conversation Monitoring

• All agent interactions are logged and auditable
• Conversation transcripts available for compliance review
• Agent responses can be configured to meet regulatory requirements

Data Retention

• Configure retention policies for conversation records
• Archive or delete conversations according to company policies
• Export conversation data for compliance reporting

---

File Attachment Support

File attachment support for Slack agents is available on the Enterprise track.

Slack users can attach files to messages the same way as any Slack message. Agents on multimodal-capable models can use the attachment content in their reply; other agents still receive the message text but do not interpret the file.

---

See Also

Add apps to your Slack workspace

Slack-side install and approval behavior