Proposed workflow. Each software feature it relies on was checked against the vendor’s own documentation on November 19, 2026. We have not run this exact setup end to end, so test it with sample data before you rely on it.
How it works
- 1. TriggerNew ticket arrivesEmail, web form or chat
- 2. AI stepAI classifies itTopic, sentiment, language
- 3. FilterRoute and prioritiseRules use the labels
- 4. You checkRight agent repliesUrgent first, right language
Software you need
-
Slack (official site, opens in a new tab) Optional
Optional alert when a very negative ticket arrives.
Estimated cost
This is the most expensive workflow in the library because it needs a Zendesk Professional-level plan and the Copilot add-on. It suits teams that already handle a steady volume of tickets.
| Software | Free plan | Paid plans from | What this workflow needs |
|---|---|---|---|
| Zendesk | No free plan used here | See Zendesk’s pricing page (we have not verified current prices) Pricing page for Zendesk (opens in a new tab) | Suite or Support Professional or above, plus the Copilot add-on to route on the labels |
Before you start
- A Zendesk Suite or Support Professional plan (or higher) and the Copilot add-on.
- An admin account, because intelligent triage is switched on through an onboarding guide.
- A short list of your real ticket topics, for example orders, returns, billing and technical issues.
- Agents or groups for each topic or language you want to route to.
Step-by-step setup
1. Switch on intelligent triage
Follow Zendesk’s onboarding guide for intelligent triage and choose which channels to classify. Email and web form are the usual starting point.
2. Review the topics it detects
Let it classify a few days of new tickets before you build rules. Compare its topic labels with what the tickets were really about.
3. Route by topic
Create a rule that sends each topic to the right group. For example, billing questions go to whoever handles payments.
4. Prioritise unhappy customers
Create a rule that raises the priority of tickets marked Negative or Very Negative, and optionally posts an alert to a Slack channel.
5. Route by language
If you support more than one language, send tickets to agents who speak it.
6. Check weekly
For the first month, review a sample of tickets each week and adjust rules where labels and reality disagree.
Test it before you rely on it
- Send test emails for each main topic from a personal address, including one clearly angry message.
- Check each ticket’s topic, sentiment and language fields.
- Confirm the routing and priority rules fired as expected.
- Check the confidence fields before routing anything important on low-confidence labels.
Troubleshooting
- Tickets are not classified. Check the channel is selected in intelligent triage settings. Zendesk classifies tickets that start with a public comment.
- Labels are often wrong. Route only on labels with high confidence, and send the rest to a general queue.
- Rules cannot see the triage fields. Using classifications in workflows needs the Copilot add-on.
- Language looks outdated. Language is not re-detected on every customer reply, so it reflects the start of the ticket.
Security and privacy
- Ticket content is processed by Zendesk’s AI features. Review Zendesk’s data processing terms against your own privacy policy.
- Avoid asking customers to email card numbers or passwords; send them to a secure form instead.
- Keep a human reply on sensitive or negative tickets rather than an automated answer.
Frequently asked questions
Can a smaller team do this more cheaply?
Yes, with simpler tools: a shared inbox with manual labels, or an AI chat tool that answers common questions first. See the customer enquiry workflow for a lower-cost option.
Does triage reply to customers?
No. It labels tickets so your rules and agents can act faster. Replies stay with your team.
Which languages does it detect?
Zendesk says about 150.
What we verified
These are the software features this workflow depends on, each checked on the vendor’s own help or pricing page. Everything else on this page, including how the steps fit together, is our proposal.
-
Zendesk intelligent triage uses AI to classify new support tickets by topic, sentiment, language and entities such as product names.
Zendesk help: intelligent triage (opens in a new tab) · checked Nov 19, 2026 -
Sentiment is one of Very Positive, Positive, Neutral, Negative or Very Negative; language detection covers about 150 languages.
Zendesk help: intelligent triage (opens in a new tab) · checked Nov 19, 2026 -
Classifications are available on Suite and Support Professional plans and above. Using them in workflows (routing and prioritisation) requires the Copilot add-on.
Zendesk help: intelligent triage (opens in a new tab) · checked Nov 19, 2026 -
By default it classifies tickets from web form, email and API, and messaging from the Web Widget, WhatsApp and Facebook Messenger.
Zendesk help: intelligent triage (opens in a new tab) · checked Nov 19, 2026
Workflow last updated . How we review


