First, you have to follow the steps 1-3 in Integrating with Zendesk to have your Zendesk integration set up.

Go to the Datasets page and select the Zendesk option.

Click add a dataset and then choose the Zendesk option in the "Tickets" category

Choose an appropriate timeframe for the dataset. Consider if you have had any major changes to the typs of tickets you receive. Ticket volumes above 50,000 tend to be unnecessarily large.

Depending on the volume of tickets you are synchronizing this step can take a couple of minutes. An email will be sent out to you when the dataset is ready.

Now its time to create a new model, first head over to the models section. Then click "New Model"

Now we click create continue to create our own model from scratch

You will now be asked to choose a dataset to train the AI model with. We pick the Zendesk dataset we just synchronized.

Now you will be asked to choose the column that contains the text that you wish to classify. This is the column with the

Next you will be asked to choose labels that you wish to train the model on, since we are going to train the model on a new field we will skip this step.

After creating the model it will appear in your models section. We click in to the model and click start training.

This will bring us to the training view. Here we will train the text classification by instructing the AI model how to categorize different support tickets. We start this by adding labels that are appropriate.

Once we have added two labels we will have to specify if the model we are creating should be a single choice model or a multiple choice model. If we are assigning a ticket field value in Zendesk you should specify the model to be Single Choice. If you are building a model to assign tags you should choose multiple choice.

Now it is time to train the model by clicking the correct label for the data points that are presented. If you need additional labels you can always click "Add a Label+"

While training the model you can head out of the training view and go to the metrics view to evaluate the models performance.

In the metrics view you can assess how well the model is performing the classification task. More about the metrics view can be found in this article.

You can also test out the models live by writing texts for it to classify in the overview section for the model

Now we head back Zendesk to activate the model

Go back to Zendesk and the Labelf app in the support view. Refresh the page to update the model list.
​

Pick the model you have just trained and the field that you trained your model on (or create a new field).

Hit activate and your model will now automatically update your incoming support tickets with the correct field! It will also set a tag with the automatic label.

If you need to change how the model interacts with your Zendesk enviroment once activated you can click "Edit Model Settings" This will take you to the trigger that is activating the model. There you can for example limit the model to only activate on a certain type of tickets.

In the trigger you can change the default conditions freely