An AI chat front door helps your team handle tough questions. When the bot is not confident, it collects an email and notifies a human in Slack. This is ideal for support teams that want fast handoffs without losing context.

The chat starts with a Chat Trigger. An AI Agent uses GPT4 with a short memory window to try to answer. If the agent is unsure, it calls a custom tool that runs a sub workflow. That sub flow checks if the user shared an email. If yes, it posts the request and email to a Slack channel and sends a confirmation back to the user. If no email is found, a Code node asks the user to repeat the question with their email so support can help.

Setup needs an OpenAI API key and a Slack bot token. You choose the Slack channel that will receive escalations. Expect faster triage, fewer dead ends, and clear ownership when the bot cannot answer. Great for website chat, internal help desks, and pilot phases while your bot is still learning.

• Chat Trigger starts a live chat session and passes user input into the flow.
• AI Agent uses the GPT4 node to draft answers and decide when it needs help.
• Window Buffer Memory keeps recent messages so the bot stays on topic.
• Custom tool dont_know_tool calls a sub workflow when the agent is unsure.
• Execute Workflow Trigger runs the sub routine that handles escalation steps.
• If node checks whether the chat includes an email address.
• Code node prompts the user to add their email when it is missing.
• Slack node posts the request and email to a chosen support channel.
• A final confirmation message tells the user a human has been notified.

• Reduce unknown question handling time from 10 minutes to 1 minute by routing directly to Slack
• Automate up to 80% of escalations when the bot is not confident
• Capture contact details before handoff to cut follow up back and forth by 60%
• Connect chat and Slack so teams work in one place
• Scale to handle more chats without adding more agents

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with Slack and OpenAI. See the Tools Required section above for links to create accounts with these services.
3. Open the GPT4 node. In the Credential field, click Create new credential, choose OpenAI API Key, then paste your key from your OpenAI account API page. Name the credential and save.
4. Test the OpenAI credential by executing the node or running a short chat in the n8n Chat panel to confirm a response is returned.
5. In Slack, create or use an app with chat:write and channels:history as needed. Install it to your workspace and copy the Bot User OAuth Token.
6. Open the Slack node Message Slack for help. In Credential to connect with, click Create new credential, choose Slack, paste the Bot token, and save. Follow on screen instructions if prompted.
7. Set the Slack channel in the Slack node to the support channel where you want alerts to appear.
8. Open the Chat Trigger node and copy the Test URL if needed. In the n8n canvas, click Chat at the bottom and send a message to start a session.
9. Send a question that the bot likely cannot answer and include an email like user@example.com. Confirm a message appears in your Slack channel and the chat replies that a human was notified.
10. Send a question without an email. Confirm the chat replies asking you to repeat the question with your email address.
11. Review the If node Check if user has provided email. Adjust the email detection rule if your format needs differ.
12. Optionally tune Window Buffer Memory to keep more or fewer recent messages depending on your chat length.
13. If Slack posts fail, verify the bot is invited to the channel and has chat:write scope. If the Chat Trigger does not respond, ensure the workflow is active and the correct URL or Chat panel is used.