How Notion AI Agents Automate Work Inside Slack (Full Transcript)

[00:00:04] Speaker 1: Notion is where you can have all your projects, documents, your wiki, all in one place.

[00:00:11] Speaker 2: When we were smaller, it worked for employees to DM me directly if they had a request. As our team grew, it just became more chaotic. How we were operating wasn't working anymore.

[00:00:24] Speaker 3: That really felt like the natural point for Notion to extend into Slack is because Slack is where so much work is done.

[00:00:31] Speaker 1: Having Slack connected to Notion is super important to have a complete picture of your work. With Notion AI, you take it a step further where you can have custom agents automate your work for you so you don't have to do it yourself. For example, answer questions in Slack that were previously answered by humans using the knowledge that's in Notion.

[00:00:52] Speaker 3: One of the main reasons why we have custom agents inside of Slack is because that's where you work, and we don't want to take you out of where you work. Integrating with Slack was really great because they make it really easy to integrate with the key primitives that make Slack Slack. Real-time collaboration is extremely important for us, and feeling like a team member is about being fast. That's why the real-time search and streaming APIs were such a great fit for our needs. But one of my favorite things in development with Slack was the typing indicator that allowed users to immediately see that our custom agent was working. Custom agents and permissions is an extremely important topic to us. We took a model where we build up the permissions from nothing. So by default, a custom agent doesn't have permission to do anything. So you can start with the agent, it can read these certain Slack channels, or it can read and reply, or it can just read and send any sort of message inside of Slack. Supporting all these different permission models was easy to do with Slack's APIs.

[00:01:48] Speaker 1: I have me an agent that updates me on my projects every week. Now the agent will start creating itself. You don't have to be an engineer to create these incredibly powerful workflows. As long as you have an idea of how the work should happen, you can just instruct that, and the agent will simply do that. So I built Smilers to answer questions about our office.

[00:02:15] Speaker 2: Smilers' knowledge base lives on a Notion page. Whenever someone puts in a request, like, where can I find the first aid station, Smilers will do his best to answer the question. If for whatever reason, Smilers gets it incorrect, our team can give him the correct answer, and Smilers will then update his own knowledge base. He can then respond with the correct answer. Having Smilers in Slack has freed up time on our team to do the work that really matters.

[00:02:46] Speaker 3: Working with agents are saving me massive amounts of time, certainly two plus hours worth of work a day.

[00:02:53] Speaker 1: It brought me back to the feeling of being a builder, and that lowers the bar of entry for everyone. I think we're honestly scratching the surface of what agents can do in Slack.