Customer Ticket Triage Agent

Scale your customer support with an agent that responds to tickets and improves automatically by learning from solved cases.

Postman Templates Illustration.

Overview

This agent responds to incoming Zendesk tickets by searching a vector database of previously resolved tickets. If a relevant match is found, it replies automatically. If not, it routes the ticket to a human agent. Once the issue is resolved, the question and answer are stored to improve future responses.

This reduces duplicate tickets and escalations by expanding the agent’s knowledge base and freeing up human teams to focus on more complex issues.

Tools used by this agent

  • Zendesk: Receives tickets, detects status updates, pulls conversations

  • Pinecone: Retrieves and stores Q&A pairs via vector search

How does this agent work?

The agent is triggered by Zendesk webhooks on ticket events:

  1. New Ticket Flow:

    • Extracts the customer question.

    • Queries a Pinecone vector database to check if a similar question has already been answered.

    • If a match is found, the agent responds to the ticket automatically. If not, it notifies a human agent to step in.

  2. Solved Ticket Flow:

    • Fetches the full conversation from Zendesk.

    • Processes the text into a clean Q&A pair

    • Stores the new pair in Pinecone with a semantic vector embedding for future lookup.

The result is an agent that improves with each question asked without requiring manual retraining or updates.

Get started fast. Clone this agent template in Postman

Use Template

How to use this template

This agent requires the following configurations to function

Zendesk

- zendesk_subdomain

- zendesk_agent_email

- zendesk_api_key

OpenAI

- openai_api_key

Slack

- slack_bot_token

- slack_channel_id

Pinecone

- pinecone_host

- pinecone_api_key

Set up the template

To set up the template, do the following:

  1. Clone the agent template from the Flows workspace.

  2. Request API keys for the following:

  3. Set up Slack:

    • Create a Slack app.

    • Get a Slack token with chat:write permissions.

    • Get the Slack channel ID.

  4. Add the following credentials to the Configurations panel to the right of the flow:

    • zendesk_subdomain

    • zendesk_agent_email

    • zendesk_api_key

    • openai_api_key

    • slack_bot_token

    • slack_channel_id

    • pinecone_host

    • pinecone_api_key

  5. Deploy the agent to the Postman cloud. Copy the action_url from the deployment output.

  6. Set up a webhook in Zendesk for ticket creation and status change updates that point to the deployed action URL. If needed, customize response logic or fallback handling.

  7. Review the output:

    • Monitor how the agent begins handling repeat questions

Popular Templates

Route urgent tickets automatically. Icon.

Route urgent tickets automatically

Detects severity of incoming tickets and escalates them to PagerDuty, Slack, or email.

Plan trips and generate itineraries. Icon.

Plan trips and generate itineraries

Builds Notion-based travel plans with activities, weather, and packing lists.

Summarize and analyze industry news. Icon.

Summarize and analyze industry news

Pulls and analyzes news articles from APIs, then sends a daily summary straight to Slack.

Score GitHub issues by urgency. Icon.

Score GitHub issues by urgency

Analyzes comments and sentiment to summarize GitHub issues and flag those needing triage.