Before You Begin: Identify the Requirements

Before you begin with Coveo Case Assist, it’s important to have all the information you need to configure the solution. Follow the steps outlined in this article to identify the requirements for your case creation experience.

Step 1: Define the Content of Your Form

The first step to creating a seamless case creation experience is to define your case creation flow. Even if you start from an existing flow, make sure to review the content of your case creation form. It’s also good practice to talk to your Support team to determine the type of information that’s considered most useful for resolving cases.

Describe problem screen

When designing your form, determine the type of information you will need the end user to provide in order to consider a case complete.

Common required inputs are:

  • Issue title or short summary

  • Issue description

  • Product or feature related to the issue

  • Issue category

  • Specific identification information (e.g., serial number)

  • File upload option

  • Priority level

Make a list of all the required inputs for your form and classify each entry as mandatory or optional. It’s also good practice to list the form elements you plan to use, as well as the character limitations (see Design Your Form).

EXAMPLE
Required information Description Mandatory or optional Form element Character limitations (if applicable)

Title

Provides a short summary of the issue.

Mandatory

Text input

255 characters

Description

Provides details about the issue.

Mandatory

Text input

32K characters

Product related to the issue

Lists products related to the issue

Mandatory

Selection

Keep in mind that the more information you request, the more effort you request from the user as well. Try to keep the required information to a minimum.

Step 2: Identify the Functionalities You Want to Use

Coveo Case Assist supports the functionalities documented in this section. Determine which one(s) you want to use.

Document Suggestion

Automatically suggests documents that are relevant to the case that’s being filled by the end user. This functionality aims to augment case deflection by potentially resolving a customer’s case before it’s submitted.

Document Suggestion

Case Classification

Automatically suggests classifications that relate to the case that’s being filled by the end user. This functionality helps end users select the right classification values when creating a case.

Case Classification

Step 3: Identify the Prerequisites for Your Configuration

Once you have determined if you want to use either one or both functionalities, it’s important to identify the prerequisites for each of the following configurations.

Document Suggestion

Follow the guidelines in this section to ensure you have the required information to configure the Document Suggestion functionality.

Define the Content to Display

When implementing the Document Suggestion functionality, it’s important to first define the content you want to display to your users. Begin by making a list.

Common content to display:

  • Public documentation of any sort (e.g., text documents, web pages, and videos)

  • Knowledge articles

  • Community posts

  • FAQs

Define the Information to Present

Next, you must define the information you want to present for each type of document.

Common information to present:

  • Title

  • Description

  • Creation or modification date

  • Document type

  • Language

  • Video duration

Make a list of all the content you want to display and of all the information you want to present.

You will need to know the name of the fields that are associated to the information you want to display. To determine the name of a field, access the Content Browser (see Review Item Properties).

EXAMPLE

To display a YouTube video or a Knowledge article, you would typically need the following information.

Content to display Information to present Field name

YouTube video

Title of video

title

Description

summary

Thumbnail

ytthumbnailurl

View count

ytviewcount

Knowledge article

Title

sftitle

Excerpt

sfsummary

Last update date

sflastmodifieddate

Author

sflastmodifiedbyid

Verify if the Index Must be Updated

From the list you created in the previous section, use the Content Browser to validate if all the information is in your Coveo index. If it isn’t, update your sources.

Refine the Items to Filter

To configure the Document Suggestion functionality, you need to create an Automatic Relevance Tuning (ART) model with the ITD option configured as specified in Configure an Automatic Relevance Tuning Model for Case Deflection or Case Assist Document Suggestions. This model must also be associated with the query pipeline dedicated to the Case Assist configuration you’re creating.

Optionally, you can also specify query expressions (using the Coveo Query Syntax) to refine the items that should be suggested by the Document Suggestion functionality (see Define the Content to Display).

Common filter examples:

  • @objecttype==Troubleshooting

  • @objecttype==Knowledge

  • @objecttype==CommunityPost

  • @objecttype==FAQ

Case Classification

Follow the guidelines in this section to ensure you have the required information to configure the Case Classification functionality.

Identify the Case Classification Method to Use

To configure Case Classification, you must first select a Case Classification method. This method determines the process that’s used to generate classification suggestions. The available options are:

  • Case Similarity (Default) identifies cases similar to the one being written based on common vocabulary. The most common classifications from similar cases are suggested as classifications for the current case (see Generate Classification Suggestions Based on Case Similarity). This method is ideal when dealing with support cases that contain very little text (8 words or less).

  • Context Recognition leverages natural language processing (NLP) techniques and a deep learning algorithm that takes into account common vocabulary, contextual nuances, and key concepts of the case description. Increasingly relevant classifications are provided over time (see Generate Classification Suggestions Based on Context Recognition). This method is ideal when dealing with larger support cases (8 words or more).

Create a Case Classification Model

If you choose to generate case classification suggestions based on Context Recognition, you must create a Case Classification Model. In order to create this model, you will need to ensure that the support cases you want to use to train the model have a sufficient amount of data in the training fields, as well as a sufficient amount of classified support cases.

You will also need to specify:

Define the Content and Fields Required to Train the Machine Learning Model

To predict categories, the Case Classification functionality learns from support cases that have been correctly classified to provide classification suggestions on cases that haven’t yet been classified.

You will need to both identify the document type that corresponds to cases in your Coveo index (e.g., sfcase for Salesforce), as well as the fields that contain the most distinctive language about a case (typically, subject, description, and resolution). This is the information that the Case Classification model will use to train.

Make a list of all the fields that need to be indexed.

Use the Properties button in the Content Browser to access the list of fields for a given document type.

EXAMPLE

To index a case in Salesforce, you will typically want to index the following fields.

Object or document type for cases Required information Field name

sfcase

Title (mandatory)

sftitle

Description (mandatory)

sfdescription

Creation date (mandatory)

date

Define the Fields Machine Learning Should Learn to Classify Cases

All categories that you want Case Classification to predict must fit the following requirements:

  • The category must correspond to a field indexed on cases in Coveo. You can use the Content Browser to validate that a field has been correctly indexed in Coveo.

  • There should be at least two possible values for a given field. Otherwise, no prediction is possible since only one answer is possible.

  • There should be at least 500 cases for each given value that you want the model to predict confidently.

Make a list of the fields you want to predict and add these to the list you started in the previous section.

EXAMPLE
Object or document type for cases Required information Field name

sfcase

Title (mandatory)

sftitle

Description (mandatory)

sfdescription

Creation date (mandatory)

date

Field to predict

sfreason

Field to predict

sftype

You should now have all the fields necessary for Case Classification to work. Make sure to index all the fields you recorded.

What's Next for Me?