Hello,
I've built and optimized an AI Slack bot for quick AI actions. It's built on top of my existing powersync architecture - which allows to connect Airtable and Slack.
I think there's an advantage in a narrow AI like this over generic AI agent with access to Airtable via MCP.
A generic AI agent utilizes the MCP to fetch more context. It might load the schema, or extra information from Slack and then start performing actions.
This AI bot has context carefully prepared ahead of time. That means it can act quicker - there's no need to rely on reasoning (which slows things down) and also it doesn't require a large model, it can get a way with smaller but more performant one.
The overall promise of such AI bot is to act quickly in Slack when needed - add tasks, manage activities, add contacts, log standup records.
The admin UI allows to define a precise boundary - which tables and fields can be accessed and if records can be created, updated or both.
Additionally - once a record is created, there's an "Open" button, which opens a form directly in Slack that allows to edit the record further.
Feel free to reach out if you're interesting in giving this a go, I can provide premium support. And I'm also very eager to hear any experience with other approaches to the same problem (OpenClaw, AI agent platforms) and what are the pros and cons there.