Intelligent triage uses AI to automatically classify new customer support tickets by topic, sentiment, language, and entities, such as product names. By incorporating these AI classifications into your workflows, you can automate repeatable requests, eliminate manual triage, guide agents in real time, and act quickly on high-risk tickets.
This article provides an overview of intelligent triage workflow examples and includes step-by-step guidance for common use cases. As you read through them, remember that you can always modify or expand on them to better support your specific needs.
This article contains the following topics:
Related articles:
Understanding intelligent triage workflows
Intelligent triage classifications — topic, sentiment, language, and entities — can be leveraged to optimize your service operations by:
- Understanding your operations: Identify exactly why your users are reaching out and how they're feeling about it. For example, track against your KPIs.
- Automating top recurring requests: Allow customers to self-serve based on topics. For example, send an auto-reply to high confidence topics.
- Eliminating manual triage: Display incoming requests to specific agents or groups. For example, assign incoming requests to the right agent or group automatically.
- Streamlining operations: Discover escalations early on. For example, customize SLA rules based on topic and sentiment.
| Workflow | Example |
|---|---|
| Route tickets | Use a trigger to route tickets with a billing topic to the billing team |
| Deflect tickets | Use a trigger to send a link to your refund policy when a refund topic is classified |
| Escalate tickets | Use an SLA policy to escalate tickets when very negative sentiment is classified |
| Adjust automated responses | Use a trigger to send an empathetic auto-reply when negative sentiment is classified |
| Route to language-specific teams | Use omnichannel routing to route tickets to a Spanish-speaking team when Spanish is classified |
| Send localized auto-replies | Use a trigger to respond in the customer's classified language |
| Auto-populate ticket fields | Configure entity extraction rules to automatically fill a product field when a product name is classified |
| Route by product or service | Use a trigger to route tickets to the relevant team when a specific product entity is classified |
You can build these workflows manually using the examples in this article, or use AI-powered recommendations to have Zendesk suggest workflows based on your account's ticket data.
Intelligent triage workflow examples
For detailed examples, see the following step-by-step workflows:
- Deflect: Redirect customers to self-serve or correct destination
- Route: Send tickets to a specific language queue from a general queue
- Reduce: Proactively request more information
- Forward: Use webhooks to pass information to external resources
- Escalate: Prioritize tickets based on customer sentiment
- Enrich: Auto-populate and route tickets using entities
Deflect: Redirect customers to self-serve or correct destination
Scenario: An end user is contacting you about a process where submitting a ticket or contacting support is not the best course of action. Instead, the required information is provided on your website or some other source, meaning the user can self-serve by following a process or other documentation that's readily available.
Examples:
- Subscription cancellations
- User profile update requests
- Refund, return policy, and warranty questions
- Job applications
Solution:
- Identify any topics that are frequently applied to the tickets submitted in the scenario described above (for example, "Cancel subscription"). If helpful, you can group multiple related topics together (such as "Refund request" and "Refund via specific channel").
- Determine the necessary confidence level that the topic must have for you to be comfortable taking automatic action on it. In other words, are false positives acceptable, where the end user can reply if they still need help?
- Create a trigger that sends an automated reply to the end user. Include instructions on how they can complete the task, a link to where they need to go in their account or app to complete the task, or a link to an existing document that answers their question.
- Leave the conversation open-ended in case the topic was accidentally mismatched, but set the ticket to a Solved status.
Route: Send tickets to a specific language queue from a general queue
Scenario: All or most end users submit tickets to a common queue, with tickets from different languages ending up in the same queue.
Examples:
- All end users use the same contact form or email address, regardless of language
- End users use the source platform in one language, but their preferred language is different from the current language of the platform or browser
Solution:
- Option 1: Create a trigger (leaving out any topic conditions) that routes tickets to the appropriate agents or group based on language (and potentially language confidence, if necessary).
- Option 2: Use other integrations to reference the ticket's language value and take an action based on that value. For example, if you currently translate macros using dynamic content, you could instead use Liquid markup to determine which language the macro should use based on the intelligent triage Language field. This approach is useful if the requester's language isn't already set in their profile (for example, if they're contacting support from an unregistered email, or in a different language than the one they have set in their profile).
Reduce: Proactively request more information
Scenario: Customers contact support, but don't include the details required to resolve the request. Agents have to reply asking for the necessary information rather than being able to solve the request during the first touch.
Examples:
- Return/replacement requests where the customer needs to provide an address
- Processes where the customer must include a purchase order or invoice number
Solution:
- Identify any topics that are frequently applied to the tickets submitted in the scenario described above (for example, "Return order"). If helpful, you can group multiple related topics together.
- Determine the necessary confidence level that the topic must have for you to be comfortable taking automatic action on it.
- Create a trigger that sends an automated reply to the end user, prompting them for the required details if they haven't already included them. This gives the customer the opportunity to reply before the agent sees the ticket, and makes it more likely that the agent can solve the ticket in a single touch.
Watch the video below for an example of how to set up intelligent triage to address this scenario.
Intelligent triage: How to proactively request more information from your users in tickets (3:10)
Forward: Use webhooks to pass information to external resources
Scenario: Customers contact support for a request that requires involvement from an external team or system. Agents must manually forward these requests to the appropriate destinations.
Examples:
- End users reach out to customer support to change their email address or other contact details, but that's owned by a team outside of Zendesk
- Certain requests require compliance or other processes to be followed outside of Zendesk
Solution:
- Identify any topics that are frequently applied to the tickets submitted in the scenario described above (for example, "Change email address"). If helpful, you can group multiple related topics together.
- Determine the necessary confidence level that the topic must have for you to be comfortable taking automatic action on it.
-
Create a trigger that sends an
automated reply that does both of the following:
- Informs the requester that their ticket has been received.
- Uses a webhook, email target, or other means from within the product to forward the relevant details from the customer's request to the appropriate external team. For example, you might send an email to an external team that includes the requester's name, email address, subject, and original message. The external team can then process the customer's request in their system.
Escalate: Prioritize tickets based on customer sentiment
Scenario: Customers contact support with a negative or very negative sentiment, indicating frustration or urgency that requires priority handling before the situation worsens.
Examples:
- Customers expressing strong dissatisfaction about a billing or account issue
- Customers threatening to cancel a subscription
- Customers who have had multiple negative interactions with support
Solution:
- Identify the sentiment values that indicate a ticket requires priority handling (for example, "Very Negative" or "Negative").
- Determine whether you want to act on sentiment alone, or combine it with a topic condition for more targeted escalations (for example, escalate tickets with a billing topic and very negative sentiment).
- Take one of the following actions:
- Create a trigger to automatically assign the ticket to a senior agent or priority group, or to send an internal notification to a manager.
- Create an SLA policy to reduce the first reply time for tickets with a very negative sentiment, ensuring they are handled before lower-priority tickets.
Enrich: Auto-populate and route tickets using entities
Scenario: Customers mention specific products, services, or key details in their tickets, but agents must manually identify this information to route or prioritize the ticket correctly.
Examples:
- Product-specific support requests where a product name appears in the ticket
- Order-related issues where an order number or invoice number is mentioned
- Location-based requests where a branch or region is referenced
Solution:
- Create an entity associated with a custom ticket field that represents the information you want to capture (for example, a drop-down field for product names).
- Configure the entity's extraction rules to automatically populate the ticket field when the entity is classified in an incoming ticket.
- Take one of the following actions:
- Create a trigger to route tickets to the appropriate agent group when a specific product entity is classified. For example, route tickets to the Camera team when the "Camera Model A" entity is classified.
- Create a view that surfaces tickets based on the populated entity field, so agents can quickly identify and act on tickets for a specific product or service.