As you already know, here at Swifteq, we do our best to answer as fast as possible to our customer needs and as such, we constantly update and improve our apps.
Here are the improvements we made this November:
Updates for our Automation Apps
Usage Reports for our Automation and Agent-assist Apps
The following apps:
Will have a new “Report” feature that delivers statistics showing the number of tasks performed. This helps you assess the impact of the automation on your team. How many tickets were auto-merged, auto-tagged or translated, saving your team precious time.
Let’s check an example showcasing how you can find and use this new feature in the Merge Duplicate Tickets app.
You can view a summary of the tickets reviewed and those automatically merged by the Merge Duplicate Tickets, helping you assess the app's performance.
At the top of the Workflows on the Configuration tab, click Reports.
A pop-up with the Usage Report for Merge Duplicate Tickets will be displayed.
It reveals the total ticket count, the number of tickets reviewed by the Merge Duplicate Tickets app, and the tickets automatically merged.
The report looks as follows:
You have the option to select a period of time from the drop-down list or set a customized interval from the calendar.
Ticket Parser Autofill Updates
Add AI to the Parser Autofill App
Using AI to Extract the Value
This new feature, as the name suggests, allows you to employ AI to automatically extract a value from a ticket comment and update a ticket field with the extracted value. You have the ability to auto extract relevant information from tickets via OpenAI's ChatGPT.
You can tell ChatGPT what the value looks like, using plain English.
First, select “Use AI to extract value”. Next, enter a value to describe "What to extract", for example, "Order Number".
The "What to Extract" value guides ChatGPT in identifying the information to retrieve. It is input as text and works alongside the specified format of the value to be extracted.
The "Format of the value to extract" guides ChatGPT on what to identify. These are essentially prompts written in plain English that describe the text and format of the desired output.
Be as detailed as possible, and you can include multiple rules if necessary. Providing an example of the expected value can also be helpful.
Note: The following example demonstrates potential outcomes. You should customize and refine the ChatGPT prompt to suit your specific requirements.
Using Placeholders
You can include placeholders such as {{ticket.created_at}} in the format for value extraction. These placeholders allow AI to interpret and extract information that may not follow a standard format.
For instance, if you need to extract the Order Date from the first customer message, and the customer might have indicated terms like “today,” “yesterday,” or “last week.” Using placeholders enables AI to parse these variations accurately.
To incorporate placeholders, simply copy them from the list of supported placeholders.
Tip: In this scenario, the Custom Field designated for the Order Date is configured with a Regex Field Type using the Regular Expression provided by Zendesk, formatted as YYYY/MM/DD.
The value extraction format includes the phrase “given that the time now is” preceding the Placeholder. This phrase is essential because, when using the API, ChatGPT does not inherently know the current time.
For this use case, this addition ensures that ChatGPT assumes the ticket creation time as the current time, enabling it to calculate relative dates effectively.
Using AI Extract with Multiple Languages
Generally, the extraction functionality works across languages, even if the prompt remains in English. If you are translating tickets (e.g., with Translate Conversations), there is no need to translate prompts in other languages.
Extract the Date of an event from a ticket comment using time placeholders
To better understand how this new feature works let’s take an example.
We want Ticket Parser Autofill to use AI to extract the date that an order was made. It may extract this info from the Subject or the First Customer Message in a ticket, and add this to a Custom Field Order Date.
First, the Custom Field for the ticket needs to be created (if necessary).
Tip: When specifying the format of the value to extract, ensure the Custom Field is configured to accept that format. For example, if the field is intended to store the Order Date, set the Field Type to Regex using the appropriate Regular Expression (e.g., YYYY/MM/DD format) as provided by Zendesk.
Next, create the workflow.
Here is a summary of how to set up the Workflow:
Name and Trigger:
Name: Enter a name, e.g., Extract Order Date with AI.
Trigger: Choose when the ticket is created or updated. You can either create a new Trigger or select an existing one that has already been set up for a Ticket Parser Autofill Workflow.
Creating the Rules for Parse and Extract:
Define parsing rules to extract values from ticket fields.
Click Add parsed value to define the rule for value extraction.
Choose which fields to extract the value from (e.g., Subject, First Customer Message).
Select Use AI to extract value.
Specify what to extract (e.g., Order Date).
Enter the Format of the value to extract as a detailed plain-English description, including any placeholders for relative dates like "today" or "yesterday" (e.g., Return the order date in the format YYYY-MM-DD).
Choosing how to Update the Ticket:
Select Set Custom Field Value and choose the custom field (e.g., Order Date) to update.
Enabling the Workflow:
Toggle the Enable/Disable setting to enable the Workflow, then click Save.
This creates the Workflow and displays it in your configuration.
AI Workflow Example:
When a ticket is created with the message "last Thursday," the Workflow triggers, and the AI extracts the Order Date as the exact date based on the email timestamp.
For instance, a client sent a message, indicating that they had placed an order "last Thursday". This email was sent on Friday, 15th of November, so the order would have been placed on Thursday, 7th of November.
The email arrives and automatically creates a ticket.
When the ticket is created, the Workflow activates, retrieves the Order Date from the message, and updates the Order Date Custom Field with the value 2024-11-07.
Help Center Manager & Translate Apps
Automatically Translate the Section and Category as Well for an Article If the Translation for That Section/Category Doesn't Exist
You have the possibility to translate Categories and Sections if they do not already exist.
This is crucial because a translated article will not display unless the corresponding translations for the Category and Section are available, even if the article is published. By default, this setting ensures that all translations are displayed.
If a translation for the Category or Section already exists, the Bulk Translate feature will not alter it. However, if the Category or Section translation in the Source article has been updated, it will not currently reflect those changes.
To identify articles missing a translation for the Category or Section, you can check the Articles page. The following example highlights articles without assigned Categories or Sections.
Reactivate the Search Results and Page After the User Returns to the Articles Page from the Editor or from a Partial Translate
Another very useful feature that we added to our Help Center Manager app is the ability to return to your initial search after you opened an article or made a modification to it.
Considering that your help center may have hundreds or even thousands of articles, first, you will search for those that interest you at a certain moment. On that selection of articles you are looking to do a variety of modifications or partial translations.
You choose the first article and modify it. And, of course, you want to return to your initial selection and move to the next article to modify. Now you have that possibility. The initial search remains automatically active, so you don’t have to go back and do your search again.
Ability to Cancel a Running Task: Translate Articles, Find And Replace, Duplicate Articles, Rollback from Zendesk, Push to Zendesk, Delete Articles in Zendesk
Now, you have the possibility to delete a task that’s running. If, for example, you want to modify or add something to the current task and need to cancel it to review your previous instructions, now you can do it faster and don’t have to wait until the current task finishes.
You have this feature available for most tasks: translate, duplicate, find and replace, etc.
Let’s look at an example. You are looking to duplicate some articles, you run the task. While it runs, a message that confirms the duplication appears, and also, in the pop-up, you will notice a button that allows you to cancel the recurrent task.
Next you will be asked to confirm you are looking to delete the task. Just click OK to continue with the Cancel Task, or click Cancel to let the Duplicate continue.
The same applies to the Translation task. You can now cancel it while it is running.
Translate Conversations
Custom Prompt for Outbound Translation
Now, the Translate function comes with a default ChatGPT Translation Prompt for Outbound Messages that you can fine-tune to match your own brand language and tone. This prompt is not applied to incoming customer messages.
To view the Translation Prompt, navigate to the Translation Conversations configuration page.
It is recommended that you only modify the Translation Prompt if you are experienced with the ChatGPT prompt code.
The System Instructions configure the translation behavior and contain the core guidelines for the translation process.
These instructions define the role as a "professional translator" and outline the rules, instructions, and constraints for ChatGPT. Also, the output must be formatted as an HTML document.
The target language is determined by the {{language}} variant, which is selected based on the language settings for the specific ticket.
The User Message provides instructions or requests for ChatGPT’s response. For further information, consult the OpenAI GPT Models documentation. Changes made to the Translation Prompt can be saved by selecting "Save Prompt."
You can restore the Swifteq defaults by clicking the Reset to Default Prompt link.
There you have it, our list of updates for the month of November 2024.
Sorin Alupoaie
Sorin Alupoaie is the founder of Swifteq, an experienced software technologist and entrepreneur, he loves shipping products that solve painful customer problems. Sorin strongly believes that any Customer Service interaction represents a huge opportunity for a business to listen and improve how they deliver value to customers.