Section 6: Sales, marketing, corporate communications and customer service

Scenario 58: Routing complaints by topic and sentiment

Level: Advanced · Tools: Zendesk Intelligent Triage, ticket fields, routing and reporting rules

Who and in what situation

Zeynep, the customer service manager at an e-commerce company in Konya, wants to separate delivery, return, payment and fault requests and route angry customers to the appropriate queue.

How it is done today

The agent fills in the topic and priority fields manually; Zeynep exports the report, groups complaints manually and forwards them to the teams.

How to do it with AI

  1. Enable Intelligent Triage classifications in Zendesk with your administrator. The system can enrich new tickets with fields such as topic or intent, sentiment, language and entities. [106]
  2. Define custom intents if the predefined intents do not suit your company. For example, you may want to track “damaged delivery” and “delayed delivery” separately. [106]
  3. Use these fields in business rules to route high-priority tickets or tickets in specific languages to the right queue. Do not let the sentiment field alone determine automatic refund or compensation decisions.
  4. Generate a trend report from ticket fields. Allow agents to correct the classification and sentiment label; review misrouted tickets using a weekly sample.

Example prompt

Review ticket classifications for {period}. Provide the most common topic, recurring product or delivery issue, high sentiment risk and suspected misrouting examples under separate headings. Rely only on information available in the ticket fields. For each recommendation, list sample ticket IDs to review; clearly state if the root cause is uncertain.

Evidence and real-world example

Zendesk explains how tickets are classified by topic, sentiment, language and entities, and how these fields are used in reports, trends, routing and priority rules. No independent company case study was found. [106]

Points to consider

Classification is not definitive truth. Carry out human corrections and sample audits because of the context and incorrect intent examples in the Zendesk support discussion. Source verification is also required for enterprise agents. [106] [108]

Sources for this scenario

  1. Automatically classifying tickets with intelligent triage (support.zendesk.com)
  2. Knowledge sources summary (learn.microsoft.com)