Slack AI Assistant Setup

Summary

The meeting provided a comprehensive step-by-step guide for building a Slack AI assistant that integrates Slack, Make, OpenAI, and Airtable.
Key decisions included using Airtable for storing messages and responses, leveraging OpenAI’s Assistant with custom instructions and vector data, and routing different Slack channels to dedicated AI assistants.
Attendees discussed automating support queries for communities, error handling, and setting up multi-assistant routing via channel-based filtering.
Action items primarily revolve around initial setup, template sharing, and optional advanced configurations.

The Slack AI assistant was built to automate support for a community that receives repeated questions.
Main goals: reduce manual team workload, provide instant responses, and enable deeper conversations for niche topics.
Key platforms used: Slack (for support channels), Make (for automations), OpenAI (for AI logic), and Airtable (for message storage and tracking).

Airtable is used to log all incoming messages and assistant responses.
Recommended to create a new base, delete default columns, and add specific fields like message ID, message, user, type, timestamp, thread timestamp, client message ID, team, channel, channel type, assistant thread ID, and assistant response.
Columns should be either single line text or long text fields as demonstrated.

Slack workspace and custom app created for the integration.
Events subscription activated; a custom webhook is set up in Make to hear Slack messages.
Slack bot permissions and channel integration steps outlined in detail.
Tested connection between Slack and Make to validate event flow.

OpenAI project and Assistant (focused on a specific API topic, e.g., ffmpeg compose) created.
System instructions and relevant API documentation uploaded to train the assistant.
Assistant can provide tailored responses and examples related to the uploaded content.

Initial step: search Airtable for existing thread (to group conversation threads properly).
Next: log each incoming message to Airtable, capturing all relevant metadata.
Slack bot replies immediately to acknowledge user messages (“Thanks, please give me a moment to respond.”).
The user query is sent to the designated OpenAI Assistant; the Assistant’s response is stored back in Airtable.
Final step: Slack bot posts the Assistant’s response in the original thread.

Automation includes error handling: if OpenAI fails, users receive a prompt to try again.
For multi-assistant routing, Slack channel IDs are used to direct queries to different Assistants (via Make routes and filters).
Variables are set for response data to manage converging outputs from multiple routes before updating Airtable and posting replies.

Use Airtable as the primary database — Chosen for flexible message tracking and ease of integration with Make.
Route Slack channels to different Assistants as needed — Enables specialized support per channel and topic.
Add error handling to the automation — Ensures users are notified promptly if issues occur, preventing automation loops.