I Made AI Agent for Slack in 5 minutes — Step-by-Step Tutorial

Use Slack. I use Slack. We all use Slack here. We use it to chat between each other. We're with potential partners asking for a quick collaboration. You know the deal. But it's always bugging me to switch from it to chat GBT or content plan back to Slack. Switch after switch. There really should be an easier way to have it all in one convenient place. And in this video, I will show you exactly how to create your own custom chatbot inside Slack that will have all the knowledge you need and plenty more. Now, here's how it usually works for me in real life. Let's say I get a message about putting an ad on the channel. Cool. Makes me happy. But I don't know where to put it, which video suits best. Usually, I would open up Notion, scroll through the ever growing list of videos, see which ones still have free slots, and then take this info and go to Slack to message back to that person. And I have to manually search for information about the tool I'm about to test and all that. That's already boring and not very efficient. Doing this once a day is easy, but five 10 times a day that adds up fast. But hey, we have chat GBT for fixing all that, right? Okay, here's a typical workflow for Chad GBT. First, it can't be connected to notion directly in any reliable way. So, instead of knowing our content plan and seeing it updated live, I have to manually create a file with all the upcoming videos, which isn't difficult to do. I can just upload a screenshot of a plan to Chad GBT and it will create a file where I can just use that screenshot as a data source for the content plan. I upload a file and simply ask Chad GBT to first find slots and videos where we don't partner up with anyone. That's usually months in advance so that I can properly test the tool and integrate it into my workflow. Then I asked Chad GBT to search for the info about the tool in question and suggest a video that'll suit the best. and only then can I take this info and go back to Slack. Again, a pretty timeconsuming workflow. Instead, I'll just create my own handy chatbot with all the knowledge and embedded into Slack so that I can do everything without leaving the app. I'll do it using Run Bear because it's simply the easiest way I found. First, I go to assistance and click add assistant right here in the dashboard. Then, I pick claude or start from scratch. doesn't really matter because I can always go back later and change the underlying model or overall approach whenever I feel like it. There are four possible sources for the assistant. Claude, OpenAI assistance, Gemini or Perplexity. For this one, I'll stick with Claude. It's a well optimized choice and it's actually recommended by Run Bear itself. Next, in the configure section, I need to give my assistant a memorable name, something like assistant ad manager. The following field is the prompt box. There's already a default prompt here that you can manually tweak, but come on, we're basically living in the future. I'll copy this entire prompt and paste it into Cad GBT, followed by a quick instruction describing exactly what I want my assistant to do. After a few iterative tweaks, I'll copy the improved prompt and paste it back into Run Bear. Okay, now we're halfway there. If I want, I can switch the model version, 3. 7, 3. 5, or even 3. 0. Personally, I suggest always choosing the newest release. It's both the most accurate and surprisingly the cheapest to run. After that comes adding a knowledge source. Notion, Google Drive, Confluence, Slack, you name it. Add in Slack or any other platform is really easy. I click on it and a familiar Slack window pops up. Here I just click allow and set the permissions, the workspace it can access, the channels it'll be available in, and the start date for that access. For more historical data, pick the furthest date you're offered. I usually add notion as well, but for this lesson, I'll just upload a single file with all the videos we have planned. And that's where the true colors of Run Bear start to show. It also has an MCP feature, which is almost like giving extra tools to the AI model, kind of like spawning little agents right inside your chat. There are loads of integrations. Shopify, HubSpot, MS Teams, YouTube, GitHub, and many, many more. Over 2500 integrations in total. That's probably more than you'll ever need, but hey, I will take the versatility any day. Pick the ones you want your assistant to access. In my case, it's Slack. Then open advanced and enable reasoning. You want the smartest model possible, right? Finally, click create. Give it a second and connect Slack once again. This time, we're not connecting it as a knowledge source, but actually as a live bot inside Slack. I give it a friendly name, paste in a token, and I'm done. To get that token, you just follow the onscreen instructions. They're interactive, so you'll always know where to click. I copy the code from Slack, paste it here, and boom, it's finished. Now I can chat with my brand new assistant directly on Slack without ever leaving the app. But I'm not done yet. To properly use Run Bear's features to their fullest, I'm going to create another type of assistant. This one based on one of my existing GBTs in Chad GBT. I've actually shown you how to create a custom GBT in one of our detailed guides, so be sure to check that out when you have a

moment. This time I'll click start from scratch and scroll a bit farther down the list. After the initial four models, you'll see four additional options. Custom GPT, OpenAI assistance, Python SDK, and Lang Serve. I suggest choosing custom GPT because it's by far the easiest to set up. For OpenAI assistance, you'd need an API key from OpenAI, and we're not digging that deep today. So, I pick custom GBT, then follow the onscreen instructions, add a link to my bookmarks, open the GBT I want in chat GBT, and simply drag and drop the bookmark onto the page. Run Bear handles all the tweaking automatically, and then just asks me for a few extra details like a name for the assistant, a prompt, and the model version to use. These fields are already populated, but you can write a new prompt if you'd like. Personally, I recommend keeping the default since it's pulled straight from the custom GPT we're referencing. After that, the process is largely the same, adding knowledge sources. As you can see, I already have some files uploaded here, and these files come straight from that GBT. So, at any moment, I can upload a few more if needed. Below that, there is a handy list of extra features our new assistant can have. Image generation, web search, and so on. Turn on everyone you think you'll use. And if you're a total geek in advanced, you can fine-tune the model even further, though I personally don't usually tinker with those settings. Next, switch to channels and once again add this assistant to Slack. You'll notice a few other options here. Discord, Teams, HubSpot, Zenesk, and more. Run Bear's assistants can live in all of these, and you can hook up the same assistant to every platform. No need to create a fresh one for each. Click, click, create a new channel, and you're done. Now, my Slack workspace has two bots, each doing slightly different jobs. The first one is a bit more limited, and the second one is more featurerich. So, if someone messages us about a partnership, all I need is the name of the tool. I'll open the first assistant and ask it to go through my content plan to find empty video slots. Then, I'll ask it to suggest a slot for the tool and give that slot a placeholder name. Because this first assistant doesn't have web browsing, I'll switch to the second one we built, the one based on a custom GBT, and ask for detailed information about the tool, what it does, key features, and so on. Finally, I copy that information and paste it back into the first bot. With this extra data, it produces a much smarter slot suggestion, three suggestions, in fact. I also ask it to draft an outline for the segment script and then generate a few matching CTAs. Armed with all that, I can return to the chat with our potential partners and have a factual conversation with CTAs ready, text drafted, and a video slot already picked. Now, you might ask, why would I need that first assistant if the second one is clearly better? Well, these are two different models, different assistants, and in my workflow, the second one does indeed work better. But that's only because my daily tasks are super basic and super simple, and they can technically all be done directly in Chad GBT. Anyway, for cases where you need deeper integration with multiple platforms, detailed data analysis, and access to a huge internal database, the First Assistant is the one to pick. It's not just a chatbot. It's pretty much a full-blown agent. There's no way around it. That's extremely cool and super useful. Adding a bot like that into Slack or Notion where every member of our team can tap into it will definitely speed up productivity across the board. With all that built-in customizability and flexibility, you can create almost anything you dream of and it will just work. No extra involvement needed. I've been testing both of my bots for a solid week now and my personal productivity has definitely gone up. Even when I open the analytics panel right after the initial testing during setup, I can see that in just nine messages, I've already saved a surprising amount of time on multitasking. Run Bear is genuinely cool and I'm absolutely sticking with it. I will leave a link in the description so you can check it out yourself. Thanks for watching and I'll see you in the next video.