Reviewing and implementing AI recommendations to automate tasks (Beta)

Add-on

Copilot

Verified AI summary ◀▼

Explore AI recommendations to automate tasks and optimize ticket workflows. Use insights on intent, sentiment, language, and entities to create triggers and autoreplies, review unused business rules, and improve auto assist procedures. You have full control over implementing these recommendations, which can enhance efficiency and reduce manual workload. Access the recommendations archive for a history of actions taken.

Location: Admin Center > AI > Overview > Overview: Copilot

The Overview: Copilot page in Admin Center provides a centralized hub to manage all Zendesk Copilot features.

As you begin to use Copilot in your account, a list of AI-powered recommendations are provided and updated weekly on this page. These recommendations provide actionable guidance on how you can automate tasks and optimize your ticket workflows by creating triggers and autoreplies based on tickets’ intent, sentiment, language, and detected entities. Recommendations also provide guidance on improving your auto assist set up and suggestions for reviewing unused business rules and macros.

This article includes these sections:

Understanding Copilot recommendations
Taking action on recommendations
Viewing the recommendations archive
Examples of recommendations

Understanding Copilot recommendations

Copilot recommendations proactively identify opportunities to automate repetitive tasks, improve ticketing workflows, clean up unused business rules, and tailor Zendesk AI to your needs.

Depending on your ticket data, your personalized recommendations may include:

Suggestions for new or updated triggers and autoreplies based on tickets’ intent, sentiment, language, and detected entities. Intelligent triage must be turned on and configured in your account to see these recommendations.

See Examples of recommendations below for more information.
Suggestions to review unused triggers, automations, and macros.
These recommendations help you manage your business rules and macros that build up over time.
Guidance for improving your auto assist setup, including recommendations for reviewing procedures with a high takeover rate.
For example, say you have a procedure where agents took over in 42% of interactions. Copilot may make a recommendation to review the procedure.

Before implementing a recommendation, you can view information and supporting insights, including:

What actions the recommendation will perform.
Recommended triggers can include actions for routing to an agent or group, or updating the ticket type, priority, form, or status. Recommended autoreplies include actions for sending automatic responses.
How your resolution time could improve, how you can reduce time spent on certain tickets, or how the recommendation will help agents.
How many tickets in a certain period of time had the detected intent, sentiment, or entity, what percentage of those tickets had a certain action taken, and how much time was spent manually performing this action on average.
When a business rule or macro was last used.

If you want to take action on a recommendation, click the link to navigate to the relevant page.

If you choose not to act on a recommendation, you can provide feedback about why you dismissed it so that your future recommendations are more accurate and useful over time. You might want to dismiss recommendations that aren’t relevant, are already automated, or that need different logic.

As an admin, you always have full control over what recommendations are implemented in your account. Any decisions and actions about recommendations require your approval. You can view all accepted and dismissed recommendations from the recommendations archive.

Taking action on recommendations

Relevant recommendations are updated weekly on the Overview: Copilot page. All Copilot customers will be able to see the Recommendations section, but you might not have recommendations.

See Examples of recommendations to view examples of the types of recommendations you may see in your account.

To take action on a recommendation

In Admin Center, click AI in the sidebar, then select Overview > Overview: Copilot.
Scroll down to the Recommendations section and click a recommendation to view more details about it.
Depending on the recommendation type, the options will differ:
- To create a recommended trigger or autoreply, click Review trigger.
  
  The Create ticket trigger page opens with prefilled trigger information.
  
  Review the prefilled information and enter a Trigger name and Trigger category. Click Create trigger.
- To review an unused business rule or macro, click the button to navigate to the relevant page.
  
  Review the unused trigger, automation, or macro and delete it if necessary.
- To improve an auto assist procedure, click Edit procedure.
  
  Edit the procedure as needed and click Publish.
When you're done, a notification appears.
Click Go back to recommendations in the notification to return to the Overview: Copilot page.
You can mark the recommendation as done to remove it from the Recommendations list.
Open the Actions menu at the bottom of the recommendation and select Mark as done

If you don’t want to take action on a recommendation, select Dismiss to remove the recommendation from the list.

You can optionally give feedback about a recommendation you dismiss.

It's important to share your feedback to help improve the accuracy and relevance of future recommendations.

Viewing the recommendations archive

Recommendations that are marked as done or are dismissed are added to your recommendations archive for you to review as needed.

Viewing your recommendations archive

In Admin Center, click AI in the sidebar, then select Overview > Overview: Copilot.
In the Recommendations section, click View archive.
Your Recommendations archive opens.
Click an archived recommendation to learn more.

Examples of recommendations

The examples in this section describe the different types of recommendations you may see in your account.

Intent-based trigger recommendation example

Recommended action	Route specific tickets to group: Support
Expected improvement	Resolution time could improve by 1h
Description and rationale	Tickets with some intents tend to be routed to the same agent group. Automate this action to reduce manual triage and help improve resolution time.
Supporting insights	167 tickets (88%) had the intent: Graduation ceremony date. Most of these tickets (16.7%) were routed to the same group: Support. On average, it took 1h to manually route each ticket. This could be reduced through automation.

Intent-based autoreply recommendation example

Recommended action	Send an autoreply to tickets with intent: Transaction failed (+4 more)
Expected improvement	Resolution time could improve by 1h
Description and rationale	Tickets with some intents tend to be answered with the same reply. Automate this action to reduce manual work and help improve resolution time.
Supporting insights	167 tickets (88%) had the intent: Transaction failed, Battery charging issue, Rate lock request accepted, How to enroll in employee benefits, Sending late arrival form. Most of these tickets (16.7%) were answered with the same reply. On average, it took 1h to manually route each ticket. This could be reduced through automation.

Sentiment-based recommendation example

Recommended action	Automatically set ticket priority to High or Urgent for tickets with negative sentiment
Expected improvement	Improve agent response times and reduce escalation risk
Description and rationale	Negative-sentiment tickets are usually urgent. By increasing ticket priority automatically, these cases can be addressed right away. It’s a way to improve customer satisfaction and prevent churn.
Supporting insights	In the last 30 days, 23 tickets (12%) were assigned negative sentiment. 28% of escalations came from delayed responses to these tickets. On average, it took 6 hours per ticket to update priority manually. This could be reduced through automation.

Entity-based recommendation example

Recommended action	Change ticket type for entity: Account error
Expected improvement	Resolution time could improve by 1h.
Description and rationale	Tickets with some entities tend to have their ticket type changed to the same one. Automate this action to reduce manual triage and help improve resolution time.
Supporting insights	167 tickets (88%) had one of these entities: Account error Most of these tickets (16.7%) had their ticket type changed to: Problem On average, it took 1h to manually triage each ticket. This could be reduced through automation.

Language-based recommendation example

Recommended action	Route specific language tickets to group: Iberia support team
Expected improvement	Resolution time could improve by 54min.
Description and rationale	Specific language tickets tend to be routed to the same group. Automate this action to reduce manual triage and help improve resolution time.
Supporting insights	4370 tickets (10%) were assigned this language: Spanish Most of these tickets (98%) were routed to the same group: Iberia support team On average, it took 54min to manually route each ticket. This could be reduced through automation.