Slack integration

This integration lets you bring any Flow into your Slack workspace, allowing you to work with AI where you already collaborate, save time, and keep everything in one place.

What can FlowHunt integration help you with:

AI Assistant Flow: Have an AI bot answer your knowledge base questions, assist you with simple tasks, and help you write better.

Customer Service Integration: Monitor chatbot conversations, get pinged when AI needs your help, and jump into any chatbot conversation right from Slack. 

Custom Tools: Build and deploy channel-specific tools to simplify daily workflows and boost productivity.

How to integrate Slack into FlowHunt

1. Go to Integrations in the main left-hand menu. 
2. Find the Slack integration and click Integrate.

3. You’ll be taken to Slack’s integration screen, informing you that Flo is requesting access to your workspace. 
4. If you have more than one workspace, use the dropdown selector in the top right corner:

5. Once you’ve picked the correct workspace, review the permissions and click Allow. 

If you want to integrate a workspace you are not the admin of, you must submit an install request to the workspace admin. 

6. You’ll get notified about a successful integration.

7. Go to your Slack workspace. You should see Flowhunt as one of the installed apps: 

Now that your account is connected, it’s time to start building the Flows you’ll use. 

Overview of Slack Components

Go back to FlowHunt and enter the Flow editor. 

You’ll notice there are two Slack-related components. The Slack Message Received and the Slack Send Message. These serve as the start and end points of Slack responses and control when and how the Flow interacts with your Slack channel. 

Slack Message Received

This component represents the start of the Slack conversation with FlowHunt. You can think of it as a trigger component. It also lets you control when, where, and how Flo’s responses are triggered. 

Component Handles

This component features three output handles that route and control the behavior after various actions in Slack. Since this component is always at the start of a Flow or Subflow, there is no input handle. 

• After Bot Stop – Controls what happens after the bot has been manually disabled via Slack. 
• After Bot Start – Controls what happens after the bot has been reenabled via Slack.  
• Plain message – Controls what happens after the Slack user (you) sends a message. 

Component Settings

• Channel – Pick the channels to which you want to add FlowHunt. 
• Workspace – Pick the workspace to which you want to add FlowHunt.

Setting up channels and workspace is necessary for the Flow to work. 

Only Trigger on Mention 

This setting controls how you trigger an answer from Flo:

• If left unchecked, Flo will respond to every single message sent in the selected channel or thread. This is great for channels used only to talk to FlowHunt, such as Customer Service Chatbot channels. You shouldn’t use this in internal populated channels where people talk to each other, as FlowHunt will annoyingly barge in on any and all conversations. 

• If checked, the bot won’t respond unless called on. You can call on Flo by using the @flowhunt mention. We recommend using this when you need quick help retrieving or summarizing information, especially in populated channels. 

Slack Send Message 

This component represents Flowhunt’s Slack messages back to you. It lets you control where, how, and to whom Flo sends answers. 

Component Handles

Slack Message – This input handle specifies what is being sent as a message. It will usually be the output of an AI Agent or AI Generator. Connecting input is necessary.

After Message Send – This output handle allows you to define further actions and behavior after FlowHunt has responded. Connecting this handle is optional.  

Component Settings 

• Format as Markdown – If checked, the messages are formatted in Markdown. If unchecked, the messages have no formatting. 
• Follow Up on Thread – Ensures that FlowHunt will keep listening and answering on conversation threads it’s been included in, for example, customer service conversations.  
• Mentions – FlowHunt will need to call on you from time to time. For example, when an customer service chatbot user requests to be transferred to a human. FlowHunt will then get your attention by using the mention you’ve selected.
• Thread Timestamp – Allows you to limit FlowHunt only to a single thread. 
• Channel – Pick the channels to which you want to add FlowHunt. 
• Workspace – Pick the workspace to which you want to add FlowHunt.

Note: You need to pick channels and workspaces in both components since there are use cases when you only use one of the components, or when you want to get answers different or multiple channels. 

Using the Slack Integration  

Various use cases come with different ways of using the Slack components. Let’s cover the two most popular use cases.

AI Assistant Flow

The first major example of Slack integration we’ll cover is the simple AI Assistant Flow. This Flow will allow you to add the Flo bot to various channels and chat with it, using it as your assistant to answer knowledge base questions or help you write better. 

The bare-bone version of this Flow only requires three components: 

• Slack Message Received
• AI Agent
• Slack Send message

1. Connect the Slack Message Received to the AI Agent using the Plain Message handle.
2. Connect Agent’s message output to the input of the Slack Send message component. 

3. Set both Slack components to the same workspace and channel. Set the other settings to your liking.
4. Don’t forget to let the Agent know its task. If you need help setting up the AI Agent, refer to this guide. 

This bare-bones Flow lets you talk to OpenAI’s GPT4o model via Slack. To make it a true AI assistant, consider adding other components such as: 

• Chat History to allow the Flow to hold on to context. 
• Other LLM components to switch from the default GPT-4o to one of the many different models FlowHunt supports.
• Tools to allow your Agent to retrieve information or perform tasks. 

Here’s an example Flow that uses the Claude Sonnet 3,5 as LLM instead of OpenAI’s models. For tools, it has access to verified information from your internal knowledge base via the Document Retriever component, but it can also search Google in real-time. Lastly, we can’t forget the chat history: 

Head to the Flow Library and get this Flow as a ready-to-go template. 

Now this bot can answer your questions using fresh information and hold a conversation about the data. 

Slack Customer Service Flow

The second important use case is connecting your AI customer service chatbot to Slack. This allows you not only to monitor all chatbot conversations but also to jump in any time and have the chatbot ping you whenever it needs your help. 

This is called human in the loop. In other words, it’s when AI lets you know what’s happening and proactively asks you to take over within the chatbot window instead of just giving the users options to contact you themselves. 

All of these scenarios require quite a few moving parts to work. You can skip all the hassle and get this Flow as a template from the Flow Library. 

We can break this Flow up into three major parts. First is building the customer service chatbot itself. Second is the triggers that let you take over via Slack. The last part is just simple quality-of-life additions. Let’s start by building the chatbot first. 

Part 1: Creating The Chatbot Subflow

1. As usual, any chatbot-related Flow starts with the user Chat Input trigger.
2. Next, we set up the escalation. Start by connecting the Escalation Gateway to Chat Input: 

3. This component is where AI decides to either take the “Human Escalation” or “Bot Escalation” route. In other words, if it has AI answer the question or asks for your help instead.
4. Continue by connecting Slack Send Message component to both routes. This component ensures that both options are reported back to the Slack channel and that you stay informed about the decision. 

5. Open the settings of the Slack Send Message and set the Workspace and Channel (or channels) in which the chatbot should be active. Select the same workspace and channels in both components: 

6. Now open the Slack Send Message component connected to human escalation. This component is where the bot notifies you it needs help. You need to set up a Mention to notify the right people. We recommend using @here, which notifies every channel member who is currently online. 
7. If you want to be notified about every user message sent, you can also set up a Mention for the Send Slack Message connected to Bot Response, too. 
8. Of course, the bot must also answer the user in the chat. That’s where an AI Agent comes into play. Connect the Bot Response handle to an AI Agent input handle: 

9. The final step is to add outputs to the Agent. We want it to respond to users in chat, but we also want to listen to the responses via Slack. That’s why we’ll connect the AI Agent to both Slack Send Message and Chat output:

10. Set up the same Workspace and Channel as everywhere else.  

We now have a simple GPT-4o Chatbot that can decide when it can’t answer and needs your help. But that’s far from a proper customer service chatbot. Let’s add all the other features of a good customer service chatbot.

Part 2: Customer Service Chatbot 

We’ll need to add a couple of components to make it a true customer service chatbot. If you have experience building in FlowHunt, this part will be easy for you.

1. Let’s take it from the top, starting with the first handle Chat History. Connect history to give your chatbot the memory of past responses. 
2. Re-check that the input is connected to Bot Response from the Escalation Gateway. 
3. If you want to use a different model, you can swap the default Chat-GPT4o by connecting one of the many other LLM components to the LLM handle. 
4. It’s time for the tools. We’ll be connecting these necessary ones: 

• Document Retriever to get real and useful information from your internal knowledge base.
• URL Retriever to let the bot crawl and grab information from URLs.
• Notification Email Tool as a fallback option. Whenever you’re not available to jump in, the bot can recommend the users leave you a message, which will get automatically sent to your inbox. 
• Human Assist Tool – As a fallback option. FlowHunt supports several popular customer service solutions. This component lets users contact human agents directly via your preferred live chat solution. 

These are the tools you should connect, but feel free to consider adding other tools, such as: 

• GoogleSearch Retriever to allow the bot to look for real-time information online. 
• YouTube Search Tool to utilize your YouTube channel in the bot’s responses. 
• Other knowledge tools to further embellish the responses. 

Here’s our agent with all the necessary Customer Service components connected: 

Now you’ve built a full customer service chatbot powered by an autonomous AI agent. The chatbot uses your information and can make escalation decisions using two fallback methods. But this is a Slack-integrated chatbot, after all, and we’re still missing the part of the Flow that lets you join the conversation. 

Part 3: Adding Human In The Loop

So far, we’ve used the Chat Input trigger. This means that we can listen in on the chat conversations in Slack, but we can’t interact with them yet. Let’s fix that:  

1. This is a separate Subflow. Find some empty space and begin with the Slack Message Received trigger. This gets triggered whenever a message is sent from Slack. In other words, whenever you as a Slack user try to interact with the FlowHunt: 

2. Start by connecting the Plain message handle to Chat Output, allowing all of your regular messages to pass through uninterrupted.  

3. The way you control if you want to jump in the conversation is by using Start Bot and Stop Bot Slack message shortcuts. You’ll notice these two options are also present on the Slack message received component.

Next, we’ll set up simple Slack messages to let you know if the bot has been stopped or started:
 

4. Take the Start Bot output handle and connect it to a Prompt component. Then, output the Promt to Slack Send Message. 
5. Do the same for Stop Bot: 

6. Open Prompt -> Edit Template and write in an arbitrary message. This only serves to let you know what is happening. In our case, the messages say “Human taking over…” and “Bot taking over…” 
7. Open Slack Send Message and set up the same Workspace and Channel as everywhere else. 
8. These messages will only show up in Slack and are only informational. The chatbot user will not see them.

Now there’s just one optional quality of life improvement left. Adding welcome messages and FAQs to guide users. 

Part 4: Adding Optional Welcome Messages And FAQs

If you use the bottom-right button to open your chatbot right now, it opens with a blank window. You can give your users something to work with right out of the gate. In this part, we’ll add welcome messages and buttons to greet the users upon opening the chat. 

1. Find a space on the canvas and start a new Subflow with the Chat Open Trigger. This means that the following messages will be triggered whenever the chat is opened. 
2. Add a Message widget and put your markdown-formatted welcome message in the Message field.
3. Connect the message to Chat Output.
4. Repeat steps 2 and 3 to add as many messages as you need. 
5. Add a Button Widget to add clickable FAQ buttons.
6. Connect the buttons to Chat Output.  
7. Repeat steps 4 and 5 to add as many clickable questions as you need. 

In our case, we’ve added one welcome message and two buttons:

In chat, they will show up like this: 

Using the Slack Customer Service Integration

It’s time to test out your Slack-integrated Customer Service Chatbot. Click the Test button in bottom right corner of the canvas and write a message. 

As always, you’ll be promptly greeted by the Chatbot: 

Now let’s switch over to Slack and see what’s happening there: 

You should be able to see both the initial user message and the Chatbot’s answer. 

If you have trouble seeing either of these messages, the most likely culprit is either leaving the “Follow up on Thread” setting unchecked in the respective Send Slack Message component or choosing an incorrect channel. 

Now let’s try to jump into the conversation. Before you send a message, make sure you’ve stopped the bot so that it doesn’t try to answer along with you. Do  this by clicking the three dots next to the las chatbot messages and clicking Stop Bot Flo: 

A short “Human taking over…” message will appear. It’s the message we’ve set up in part 3. Now you can answe freely without the chatbot interfering: 

Notice how the “Human taking over…” message only appears in Slack, but not in the Chatbot window. This message is only for your convenience, informing you about the bot’s state. 

If, at any time during the conversation, you’d like the bot to take back over, simply click the Sart Bot Flo button. You’ll see the “Bot taking over…” message. However, the bot won’t say anything in the chat unless there’s a user message to respond to: