Automate Customer Support Issue Resolution using AI Text Classifier n8n Workflow Template

Try It Out!

This n8n template is designed to assist and improve customer support team member capacity by automating the resolution of long-lived and forgotten JIRA issues.

• Schedule Trigger runs daily to check for long-lived unresolved issues and imports them into the workflow.
• Each Issue is handled as a separate subworkflow by using an execute workflow node. This allows parallel processing.
• A report is generated from the issue using its comment history allowing the issue to be classified by AI - determining the state and progress of the issue.
• If determined to be resolved, sentiment analysis is performed to track customer satisfaction. If negative, a slack message is sent to escalate, otherwise the issue is closed automatically.
• If no response has been initiated, an AI agent will attempt to search and resolve the issue itself using similar resolved issues or from the notion database. If a solution is found, it is posted to the issue and closed.
• If the issue is blocked and waiting for responses, then a reminder message is added.

Need Help?

Join the Discord or ask in the Forum!

1. Search For Unresolved Long-lived JIRA Issues

Learn more about the JIRA node

In this demonstration, we'll define "long-lived" as any issue which is unresolved after 7 days. Adjust to fit your own criteria.

We'll also use the Execute Workflow node to run the issues separate in parallel. This is a performance optimisation and if not required, the alternative is to use a loop node instead.

2. Retrieves and Combine JIRA Issue Comments

Learn more about the JIRA node

To provide the necessary information for our AI agents, we'll fetch and combine all the issue's comments along with our issue. This gives a accurate history of the the issues progress (or lack thereof!).

3. Classify the Current State of the Issue

Learn more about the Text Classifier node

Today's AI/LLMs are well suited for solving contextual problems like determining issue state. Here, we can use the text classifier node to analyse the issue as a whole to determine our next move. Almost like a really, really smart Switch node!

There are 3 branches we want to take: Check if a resolution was reached, blocked issues and auto-resolving when no team member has yet to respond.

4. Sentiment Analysis on Issue Resolution

Read more about the Sentiment Analysis node

The Sentiment Analysis node is a convenient method of assessing customer satisfaction from resolved issues. Here, when resolution is detected as positive, we can ask use the opportunity to capitalise of the favourable experience which in this example, is to ask for a review. In the opposite vein, if the exchange has been negative, we can escalate in an attempt to improve the situation before closing the ticket.

AI can equip teams to provide unrivalled customer support which can differentiate themselves significantly against the competition.

6. Notify for Unanswered Questions or Response Waiting

Read more about the Basic LLM Chain node

In this step, where signals indicate that the issue is not yet ready to be close, we can try to re-engage issue participants by summarize the conversation so far and sending a reminder comment for any pending actions that were requested. This action can help reduce the number of issues which linger for too long.

5. Attempt to Resolve The Issue With KnowledgeBase

Read more about the AI Agent node

When the issue is unaddressed, we can attempt to resolve the issue automatically using AI. Here an AI agent can easily be deployed with access to knowledge tools to research and generate solutions for the user. Since n8n v1.62.1, AI Tools Agents can attach nodes directly as tools providing a very easy way to linking documents to the LLM.

Here, we use both the JIRA tool to search for similar issues and the notion tool to query for product pages. If a solution can be generated, we create a new comment with the solution and attach it to the issue. If not, then we can leave a simple message notifying the user that we could not do so. Finally, we close the issue as no further action can likely be taken in this case.