In intelligent triage part 1, we learned what intelligent triage is, and why it’s so powerful. In this article, we’re going to cover how to set up intelligent triage and reap the benefits of this feature in no time.

For our example, let’s imagine that we want Zendesk to intelligently triage any ticket with negative sentiment plus the keywords “twitter” or “tweet” to our escalations team. This way, if someone tells our customer service team they’re going to share their great experience with their followers on Twitter, the ticket will be handled as normal.

If the reverse is true, however, and the customer is threatening to tweet about their negative experience, intelligent triage can ensure that the ticket gets straight to the escalations team without wasting any time waiting to be manually triaged by a first-line agent.

Trigger Creation

To begin, head over to your Zendesk Admin centre, at YOURDOMAIN.Zendesk.com, or by clicking the four squares in the top right of any Zendesk page, then Admin Centre

‍

From here, navigate or search to the Triggers page at YOURDOMAIN.zendesk.com/admin/objects-rules/rules/triggers

In the top right of that page, click Add trigger

‍

Give your trigger a name and assign it to a trigger category. You can pick anything you like for these two fields, they’re not too important.

Let’s go ahead and add our first two conditions under the ALL section.

First of all, we only want Zendesk to apply this trigger if it picks up on something the customer has said, rather than something our agents have said in response.

To do that, create the first condition that says “Ticket is Updated”.

Follow that with another condition which states that “Current user is (end-user)”.

This means that the trigger will only fire when there is an update from an end-user, rather than an agent or admin.

Perfect, now we’ve told the trigger which messages we want it to look for, we can move on to telling it what to look for within those messages.

Picking out negative sentiment

To tell Zendesk to look for customer messages with negative sentiment, we can use intelligent triage’s new “Sentiment” attribute.

‍

When we select this, we have a choice to make. As you can see, Zendesk offers us two options for sentiments which it deems negative; Negative and Very negative.

Ideally, we’d like it to look for tickets that have either Negative or Very negative sentiment.

However, as we’re in the ALL conditions section currently, if we create one condition here for Negative sentiment, and another for Very Negative sentiment, Zendesk will look for tickets that meet both these criteria, which will be precisely zero, as each ticket can only have one Sentiment value (a ticket can’t have both “Negative” sentiment and “Very negative” sentiment.

Instead, we need to use the ANY conditions section, to specify that Zendesk should look for tickets with either Negative sentiment or Very negative sentiment.

‍If you’re feeling a bit lost by which condition to put where in your trigger, have a look through my previous article on this topic, Zendesk Triggers and Automations: when to use ALL vs. ANY

Go ahead and add these two Sentiment conditions to the ANY section of your trigger. It should now look like this:

Checking sentiment confidence

The next step is to ensure that Zendesk only picks out those tickets for which it is highly confident that the sentiment it has assigned the ticket is  correct.

After all, this is an AI estimating the sentiment of these tickets, and it can make mistakes. The great thing about tools like this is that it knows how likely it is to be correct.

With this in mind, and not wanting to flood our typically small escalations team with every message about Twitter, let’s tell our trigger to only include tickets for which it is highly confident that the sentiment is either Negative or Very negative.

Add this to the ALL section of your trigger, as we want this condition to always have to be true in order for the trigger to fire.

Adding keywords

Our final condition is to tell Zendek to only pick out tickets which are to do with Twitter. Given the other conditions we’ve already added, that would mean it would look exclusively for tickets that mention Twitter or tweeting, are from a customer’s message, and which the AI is quite certain are written with a negative sentiment.

Of course, you could simply finish off your trigger at this stage, without the Twitter stipulation. That would still send all customer messages that are highly likely to be negative to your escalations team, which some admins may find preferable.

In our case though, let’s finish off our example and make sure that intelligent triage only sends negative messages that are regarding Twitter to our escalations team.

Go ahead and use the “Comment text” attribute like below. Add it to the ALL section again.

Feel free to edit the keywords as you see fit. I’ve added “x.com”, Twitter’s new name, but not simply “x” due to concerns that the letter “x” is used relatively frequently as a standard letter in some unrelated words.

Your intelligent triage trigger’s conditions should now look something  like this:

Setting up an action

The final piece of the puzzle is to tell Zendesk what to do with tickets that meet these criteria. This may vary depending on how your escalations team is separated in Zendesk, but generally it will be either changing the tickets group or adding a tag if your escalation view relies on escalated tickets having a unique tag.

In our case, we’re going to assume that the Escalations team are reached simply by assigning the ticket to their group.

This will ensure that all agents in the L2 - Escalations group will have access to that ticket and can respond right away.

And that’s it! You are now harnessing the power of AI to automatically send troublesome customer messages directly and instantly to your escalations team, without any time being wasted in the triage stage.

I encourage you to follow along with these steps the first time and then play around with the variables as you see fit.

What other use cases can you think of to make the most of intelligent triage? Although this example may well be useful, it’s only the beginning of what’s possible!