Shelf Answer Assist End-to-End Guide



Document Version History

Version number

Modified by

Modifications made

Date modified

Status

1.0

Shelf

Initial release of the document.

26 October  2022

Expired

1.1

Shelf

Version of the guide was updated to absorb new structure of the document and the Answer Assist configuration and use flow.

11 November 2022

Expired

1.2

Shelf

New section was added on the IVR configuration via context mappings. Figures were renumbered accordingly to accommodate new imagery.

27 January 2023

Expired

1.3

Shelf

Enabling and configuring the Artificial Intelligence Suggestions section was expanded with a subsection about scoreAttributes parameter and its configuration.

1 February 2023

Expired

1.4

Shelf

Multiple amendments and additions made to the document. New chapter added to describe the new Shelf Search Suggestions feature. New chapters added to describe the Interaction History feature.

13 June 2024

Expired

1.5ShelfThe guide has been updated due to the deprecation of the AI Suggestions feature and implementation of the Real-Time Suggestions feature. Multiple changes to text and imagery have been made throughout the document.16 November 2024Expired
1.6ShelfThe guide has been updated to reflect the new capabilities implemented in Shelf Answer Assist: hiding individual content mapping filter fields from the Suggestions tab, and disabling and enabling the ability to open Search Results and Gems from Answer Assist Widget in Shelf KMS.20 December 2024Active


Document Purpose

This document has been developed specifically for the customers of Shelf’s Answer Assist to:

  • Give the users of the Shelf Answer Assist product (hereinafter - AA or the Product)  an understanding of its capabilities so that they can utilize the Product to meet their business needs.

  • Provide the AA users with information on how to enable its various features.


This End-to-End Guide covers all aspects of the Product interface, features, and capabilities, as well as the generic - suitable for all integrations - configuration settings. Details on the configuration of customer-specific integrations are laid down in the respective integration-specific Shelf Answer Assist installation and configuration guides to be provided to customers along with this End-to-End Guide.

Once any new feature has been implemented in the Product, this Guide is subject to change without notice to include new descriptions and imagery.


If you require further assistance, please do not hesitate to contact us at support@shelf.io or from the in-app chat support within the Shelf platform.








1. Answer Assist and Its Purpose

Answer Assist is one of Shelf’s Answer Automation products designed and developed to help contact centers and organizations quickly and reliably deliver accurate answers to customer questions. 

Answer Assist works as an on-demand online reference library, 24/7/365, where agents can easily access the necessary information (or content) to help customers resolve their issues and get answers to their questions.


It can be integrated with your contact center’s web-enabled solution that supports third party plugins. 


Featuring built-in AI, context suggestions, federated search and content feedback capabilities, Shelf Answer Assist:

  • Increases agents’ efficiency and productivity

Search suggestions featured in Answer Assist help agents find precise answers and surface them to customers directly in the interface of the system where it is integrated

  • Optimizes admin effort and service quality

Thanks to a simple yet robust Answer Assist configuration, Shelf makes it easier for your administrative team to help you help your organization’s customers in a faster and more efficient manner

  • Ensures long-term customer satisfaction

Shelf’s extensibility enables a more complete understanding of agent performance, from knowledge effectiveness to overall customer interactions, facilitating continuous improvement 


With its proprietary suggestions and content mapping configurations, Shelf Answer Assist finds suggestions by reading customer messages, identifying their intent, and offering relevant answers for agents directly in their work tool.



What Makes Shelf Answer Assist Different

The Shelf Answer Assist modular structure focuses on giving you only what you need and when you need it. Agents will find the features included in Answer Assist are full of the necessary tools for their interactions. Overall, agents will save time as all the unnecessary options are eliminated and they are offered only those capabilities they need to resolve issues

By default, your agents can enjoy the Search and Gem modules of the Shelf Answer Assist out of the box. The Suggestions and Favorites modules are also available from the very beginning but they need to be switched on in the Shelf Admin Panel. 

The other modules - Feedback, Announcements, and Links - are enabled on the Shelf side upon request following which your organization’s admin can switch these on for the needed accounts.

Below is a short summary of the Shelf Answer Assist modules and their purpose.

  1. Suggestions: The key functionality of Answer Assist that provides your agents with a quick and easy way to answer your customers’ questions.

  2. Favorites: Agents can save specific content and frequently used content to Favorites. This functionality offers faster and easier access to the content agents use everyday.

  3. Search: Another fundamental feature of the Shelf Answer Assist; it identifies specific keywords and phrases within documents, PDFs, Word, PPT, Excel, images, Wiki pages, and Articles; your agents can search content by various and multiple filters such as date added, recently changed, tags, categories, content type, file type, favorites, and ratings.

  4. Feedback: There are two types of Feedback in Answer Assist - feedback on content and feedback on suggestions. Feedback on content allows using your agents’ knowledge of the content efficiency. Their Feedback is critical to keeping content accurate and up to date. Agents can submit content feedback directly from Answer Assist, so content managers can quickly implement recommended improvements and updates. Meanwhile, feedback on suggestions enables Shelf to improve the accuracy and relevance of suggestions via your agents’ feedback on suggestions they receive. A quick upvote or downvote helps improve the algorithm to make better suggestions over time.

  5. Announcements: Through this functionality, your agents can be notified of the important content changes or updates. This feature can be also used to ensure and track compliance of your agents with your organization policies. Content managers can see which users receive and acknowledge specific updates. Agents can access Announcements and confirm they have read important information directly inside the Answer Assist widget.

  6. Links: This functionality lets your agents link content they used to answer customers’ questions to interactions with such customers. It might be needed if the compliance or any other regulation or policy of your organization demands agents to specify what exactly is used to respond to customer requests. 

Enabling or disabling a certain module is implemented on the Shelf side and affects the user interface of the Shelf Answer Assist displayed to your content managers and agents. Each of the above functionalities is explained in detail in the sections below.





2. Content and Context in Answer Assist

Before you proceed with making yourself familiar with other details on Answer Assist, you need to get an understanding of what the Content, Context and Content Mapping mean in terms of Answer Assist and why Shelf needs them to enable effective operation of the Answer Assist product.



      

2.1 Content

In Shelf’s terminology, content is any Gem created or stored on the Shelf Platform. It can be of various types. See the figure below for the types of content in Shelf.


Figure 1. Gem types in Shelf

In brief, having an account on the Shelf Platform, you can create any content that suits your business needs. It can be a document, a graphic, or even something like a Wikipedia page. 

The Shelf Answer Assist is designed and developed to suggest and deliver the needed content to you and your agents in order to enable prompt and grounded support for your customers. 
However, in order to let Answer Assist perform that content suggestion and delivery in a right and accurate manner, you need to make sure that you have properly categorized your content, identified relevant context, and performed all the needed context mapping settings. 

The Shelf Platform provides multiple content categorization options, which are:
  • Gem type: in addition to the Gem types shown in the figure above, there is also the Custom type of Gems that can be configured to your specific needs, and the other two Gem types which are enabled upon request by Shelf administrators - FAQ and Decision Tree. Once these two types of Gems are enabled for your organization, they can be enabled for the needed users - content managers or agents.
  • Gem category: this option allows you to assign any needed Gem any needed category(-ies) for a better grouping, filtering, or classification
    Figure 2. Viewing and managing Gem categories in Shelf
  • Gem tag: this option helps you better cover use cases for your content and allows you to assign any needed Gem any needed tag(s).
    Figure 3. Viewing and managing Gem tags in Shelf
     


2.2 Context

Context is a whole situation, background, or environment that is relevant to a particular interaction, topic, question, or answer.

For example, if a customer during their interaction with an agent speaks about an unrefunded payment, the context here may be payment system failure, bank issues, or wrong bank details indicated in a refund request. There also can be other data that can serve as an additional context and have certain importance for your organization. This data can include such things as:
  • your customer’s country, because different countries have different conditions on payments or refunds,
  • your  customer’s segment, because you can group your customers by their importance for your organization’s business and establish different request handling procedures for different customer groups,
  • your business segment, because you can establish different request handling procedures for different business units of your organization.
  • any other classification or categorization data you treat as important for distinguishing your customers or their domains.

In terms of Answer Assist, context and its mapping are required to ensure that the proper content (Gem(s)) is always ready at your agent’s hand to address the most typical problems grouped by context. 

Details on configuring the content mappings and context can be found in the relevant sections below.

Once you have made yourself familiar with this introductory data on the Shelf Answer Assist and on the Shelf understanding of content and context, it is time to learn about the standard flow of configuring and working with the Shelf Answer Assist.






3. Shelf Answer Assist Use Flow

Before you start using the Shelf Answer Assist, you need to install it on a platform where you have arranged your customer support operations.
If you need any help with your Answer Assist integration installation, please contact us at support@shelf.io or use the in-app chat within the Shelf platform.


3.1. Preparing the Shelf Answer Assist for Use

Once Shelf’s Answer Assist has been installed on your platform, you need to proceed to the global configurations on the Shelf side, which are described in the below subsections.


3.1.1. Shelf Admin Panel

Once the Shelf Answer Assist feature has been enabled for your organization, you can now access and start configuring it in the Shelf Admin Panel. For this purpose, you need to log in to your Shelf account with the Admin credentials.

Once logged in, go to the Admin Panel. 

Figure 4. Accessing the Admin Panel in Shelf

In the Admin Panel, find and select the Answer Assist Configurator option in the sidebar panel.

Figure 5. Accessing the Answer Assist Configuration menu

Following that you are taken to the Answer Assist Configurator page where you can start creating, editing, and otherwise managing configurations for the Shelf Answer Assist.

Answer Assist Configurator Settings

Figure 6. Viewing the Answer Assist configuration(s)

            
Creating and Deleting the Answer Assist Configuration 
If you have no configurations for your Answer Assist instance yet, click the ADD CONFIGURATION link button. That directs you to the new window where you can start setting your Answer Assist configuration. Up to 20 configurations can be created and configured for one account.

Figure 7. Starting the Answer Assist configuration procedure

By default, the window appears with the General tab (1) active that shows the form where you can enter the name (2) for your configuration, and its short description (3). You can also see and copy the identifier (4) for the configuration you are creating, and save it for further use.

Once all the needed information has been entered, you can click the SAVE button to save the configuration and proceed to its further setting. The DISCARD button at the bottom of the form can be used to clear the description or the name of the configuration, depending on what you have already entered in this form. The DELETE button lets you delete either the existing configuration you have opened for viewing or the configuration you are creating now. If you want to remove the configuration, click this button and then confirm your choice in the popup window.

Figure 8. Confirming the Answer Assist configuration deletion

Once the SAVE button is clicked, the configuration is saved and you can continue its further setting.


Enabling and Disabling Suggestion Sources for Answer Assist

📢Please be informed about an important change regarding one of our product features. After careful consideration and evaluation, we have decided to remove the AI Suggestions feature from our functionality.
We understand that this may impact your workflow. However, we have introduced a new feature - Real-Time Suggestions - that leverages our advanced search and AI capabilities to assist agents in handling interactions with high efficiency. Read more details about enabling, configuring, and using this feature in the respective sections below.

Enabling Features and Tabs for Answer Assist

As mentioned in the Answer Assist’s Difference section above, the Shelf Answer Assist has a modular structure and is supplied with a certain set of pre-enabled features - Search and Gem. We have discussed above how the Suggestions feature can be enabled. At the same time, the Feedback, Favorites, and Announcements features can be easily enabled via the following Settings menu of Shelf’s Answer Assist Configurator.

Figure 11. Enabling the Feedback feature and the Announcements and Favorites tabs for Answer Assist

The last two features - Links and Search Suggestions - need to be enabled via the Advanced Settings tab of the Answer Assist Configurator. See details on how to do it in the Enabling the Gem Linking (Links) feature and the Enabling and Configuring Search Suggestions for Answer Assist sections below.

Refer to the feature-specific section of this guide for more information.
The Advanced tab of the Answer Assist Configurator is the place where you can configure various Answer Assist-related settings, for example - set the default order of tabs in the Answer Assist widget, indicate which tab needs to open by default, change the order of tabs etc.

All these settings are performed by means of JSON code inserted into the Code Editor field. 


               

Setting and changing the order of the Answer Assist widget tabs

The default order in which the Answer Assist tabs are displayed in the widget is Suggestions | Favorites | Search | Gem | Announcements | Links | Settings. The respective JSON code is as follows:

{ 
  
"tabs": {
      
"order": [ 
        
"suggestions" , 
        
"favorites" , 
        
"search" , 
        
"gem" , 
        
"announcements" , 
        
"links" , 

        "settings"
        
] 
    
}
  
}

In the Code Editor of the Advanced tab, it looks as follows.

Figure 12. JSON code for the default order of tabs in Answer Assist

Figure 13. Viewing the default order of tabs in the Answer Assist widget

Legend of the tab order as shown in Figure 13
  1. - Suggestions tab
  2. - Favorites tab
  3. - Search tab
  4. - Gem tab
  5. - Announcements tab
  6. - Links tab
  7. - Personal settings tab

If you need to change the order of tabs, switch the tabs in the code as needed and click SAVE as shown in the figure below.

Figure 14. Changing the order of tabs in the Answer Assist widget

Figure 15. Verifying the changed order of tabs in the Answer Assist widget

Legend of the tab order as shown in Figure 15
  1. - Gem tab
  2. - Suggestions tab
  3. - Announcements tab
  4. - Favorites tab
  5. - Search tab
  6. - Links tab
  7. - Personal settings tab


                 

Setting the default opening Answer Assist widget tab 

By default, the Shelf Answer Assist widget starts with the Search tab open.

Figure 16. Search as the default opening tab in the Answer Assist widget

If you want to change it, you need to add the respective piece of code (see the figure below) to your JSON in the Advanced tab of the Answer Assist Configurator and save changes.

Figure 17. Changing the default opening tab in the Answer Assist widget

Figure 18. Links as the default opening tab in the Answer Assist widget


                    

Enabling the Gem Linking (Links) feature

The feature shown in Figure 18 above - Gem Linking or Links - is enabled by way of adding the respective string in the Code Editor in the Advanced tab of the Answer Assist Configurator. 

To enable it, add the following string to your JSON:

"contentLinking": {
   
"enabled": true 

}


Figure 19. Enabling the Links feature for the Answer Assist widget

If you do not add the highlighted string in the Code Editor, the Answer Assist widget will contain no Links tab, even if you have previously set it as the default opening tab.

Figure 20. Viewing the Answer Assist widget without Links tab

Refer to the Gem Linkage Functionality in Answer Assist section for more information on this feature and its usage.


Enabling and Disabling opening Answer Assist widget’s Search Results and Gems in Shelf

By default, when you work in the Answer Assist widget, you can look up for anything in the Search tab and then open, both the Search Results and the individual content items (Gems) as a result of such a search, in Shelf KMS. 


For example, we have searched and found content items by the car keyword. The results are displayed in the tab, and we can open this entire search in Shelf KMS, by clicking the Open Search in Shelf icon as shown in the figure below.
Figure 21.  Running a search in Answer Assist widget and then opening it in Shelf KMS

Figure 22. Viewing AA-run Search in Shelf KMS

We can then open any found or any other content item (Gem) in the Gem tab and, from that tab, open it further in Shelf KMS, by clicking the Open in Shelf icon as shown below.
Figure 23. Viewing Gem in Answer Assist widget and starting its opening in Shelf KMS

Figure 24. Viewing Gem, found or suggested in Answer Assist widget, in Shelf KMS


However, there can be situations when occasional clicks can occur during the agents’ conversations with customers and as a result, the important information can be closed or disappear from the agents’ screens. To mitigate this risk and to enhance agents’ focus on their current interactions with and questions from customers, Shelf has implemented the capability that allows disabling the ability to open Search Results and individual Gems in Shelf KMS from Answer Assist.


This capability is enabled via adding a dedicated parameter and its value to the JSON in the Advanced tab for the needed Answer Assist configuration. 


The parameter and its value are as follows:

Figure 25. Disabling Search and Gem opening from AA widget in Shelf KMS

The parameter is added to JSON and the changes to the needed Answer Assist configuration are saved, we can go back to the Answer Assist widget and check whether the Open Search in Shelf and Open in Shelf icons are displayed there.
Figure 26. Verifying the capability to open Search Results from AA in Shelf KMS is disabled

Figure 27. Verifying the capability to open Gems from AA in Shelf KMS is disabled


Enabling and Configuring Real-Time Suggestions

By default, the Real-Time Suggestions feature is disabled in the Shelf Answer Assist and needs to be enabled manually. To do so, go to the Shelf Admin Panel, navigate to the Answer Assist Configurator by selecting this option in the sidebar panel, open your Answer Assist configuration by selecting it from the list of available configurations, and then open the Suggestion Sources tab. Once in this tab, switch the Shelf Real-Time Suggestions toggle on and confirm your choice by clicking the SAVE button. The Any option is the default value, which includes all available libraries, but you can narrow down the filter to the appropriate library(-ies). Multichoice is supported and you can select up to 100 libraries.

Figure 28. Enabling the Real-Time Suggestions feature for Answer Assist


Now, if you need to enable additional settings for real-time suggestions to be generated and displayed, go to the Advanced tab and make the relevant changes to the JSON code there.


In particular, you can tell your Shelf instance and Answer Assist what other settings need to be accounted for by real-time suggestions.

Let’s discuss these parameters in detail in the following table.

realTimeSuggestions

This is the parameter that enables additional settings for Real-Time Suggestions. It requires no further values.

enablePIIRedaction

A subordinate optional parameter that can attain true or false value; by default, if enabled, it has true value, which means that personal or sensitive information, such as social security numbers, bank cards, phone numbers, home addresses and alike, is redacted in accordance with the Shelf PII handling policy. If customers want to use their own PIIRedaction methods for their transcripts, they need to set this parameter to false.

experiments

This optional parameter is subordinate to realTimeSuggestions and opens a set of other parameters that let users configure suggestion language and message classification.

suggestionsLanguage

The language of suggestions is defined automatically by default. However, using this parameter, customers can leverage what suggestions they are shown and narrow them down to the specific language only.

skipClassification

Another optional parameter subordinate to experiments. It is false by default; if it's set to true, the messages are not classified by whether they should trigger suggestions or not.

filtersMapping

This optional parameter opens a set of subordinate parameters which are used by both Real-Time Suggestions and Search Suggestions. Once enabled, it allows additional filters, such as collectionIds, categories, and tags to be passed from external services as context.

collectionIds

This optional parameter is enabled by filtersMapping and is arbitrary or defined by the external system name of the field in the external context.

categories

This optional parameter is enabled by filtersMapping and is arbitrary or defined by the external system name of the field in the external context.

tags

This optional parameter is enabled by filtersMapping and is arbitrary or defined by the external system name of the field in the external context.


Once you have enabled Real-Time Suggestions in the Suggestion Sources tab and, optionally, configured optional filtering and other parameters via JSON in the Advanced tab, you can expect to receive real-time suggestions in your widget’s Suggestions tab. More details on the actual operation of the feature in the Answer Assist widget can be found in the respective chapterReceiving Real-Time Context-Mapped Suggestionsbelow. 


                

Enabling and Configuring the Context-Mapped Suggestions

Context-mapped suggestions are present in Answer Assist by default but for their proper operation you need to activate and configure this feature via the below two-stage procedure.

The first stage requires to perform content mapping, and the second stage anticipates configuring context mappings by means of JSON code. 


Stage 1: Content Mapping in Shelf

You need to set up the content mapping in order to allow the Shelf Answer Assist to provide proper suggestions to your agents during conversations with customers.

To do so, follow the below steps:
  1. Once logged in to your Shelf account with the Admin role credentials, go to the Admin Panel and access the Content Mappingsfeature and page, respectively, there.
    Figure 29. Accessing the Shelf Admin Panel 

    Figure 30. Accessing the Content Mappings feature in Shelf

  2. Once on the Content Mappings page, create new mappings for your content (Gems) either (1) manually by clicking the ADD MAPPING button and then populating the respective cells in the Content Mappings table grid, or (2) by uploading a prepared CSV file shown below via the UPLOAD CSV button.

Figure 31. Preparing the CSV file for content mappings

Figure 32. Adding new content mappings, manually or via CSV upload

When configuring content mappings for the Shelf Answer Assist’s context-mapped suggestions, make sure to properly enter correct mapping values, in accordance with the table below.

Table 1. Content mapping values in Shelf


Field

Type

Description

Example

intent

string

(required 1 - 256 characters)

Uniq code for the system combination of the intentGroupId and intent

Breeds

intentGroupId

string

(required 1 - 256 characters)

Useful to group several intents by some criteria (eg by market, specific topic, interaction channel etc.)

dogs

mappingType

enum: CONTENT ITEM | SEARCH

Auto-populated based on the mappingValue

mappingValue

URL (string or array of strings split by ,)

URL of search results, single gem or list of gems for an Intent

https://domain.shelf.io/?cp=1&fi=identifier&it=card&ps=20&sb=CREATED_DATE&sl=en&so=DESC&tg={{breed}} (exemplary Search results link)


If you want to get context suggestions from a wide set of content items (Gems), you need to conduct a search by a needed keyword in Shelf and then copy the URL of the Search Results page from the address bar of your browser and insert it into the respective mappingValue field.

Figure 33. Shelf Search Results page explained

If you want to narrow the suggestions to some specific content (Gem(s)) only, you can apply some filter to return the list of Gems that will be mapped to the Intent and copy the search results URL in from your browser search bar.

Figure 34. Viewing the filtered search for content in Shelf

Finally, you can narrow the search and, respectively, suggestions to a specific Gem only. For this purpose, open the Gem page in Shelf and copy its URL from the address bar of your browser and use it as a mappingValue for the needed Intent.

Figure 35. Viewing the Gem page to be used for the Intent

Figure 36. Saving the new content mapping

3. Save the prepared content mapping by clicking the respective - SAVE - button.

Once the above steps are accomplished, you can proceed to the second stage, that is to configure the context-mapped suggestions settings.

Stage 2: Setting up the Context-Mapped Suggestions in Shelf

To perform it, follow the below steps.
  1. Go to the Admin Panel and navigate to the Answer Assist Configurator feature. 

    Figure 37. Accessing the Answer Assist configurations

  2. In the Answer Assist Configurator window, open the needed Answer Assist configuration, navigate to the Advancedtab and add the JSON configuration for your context-mapped suggestions in Answer Assist integrated into your platform.

    Figure 38. Accessing the Advanced tab in Answer Assist Configurator in Shelf

Exemplary JSON configuration

{
  
"contextSuggestions": {
    
"contextTypes": [
      
{
        
"accessor": "dogBreedStr",
        
"options": [
          
{
            
"value": "Large breeds",
            
"displayName": "Large breeds"
          
},
          
{
            
"value": "Medium breeds",
            
"displayName": "Medium breeds"
          
},
          
{
            
"value": "Small breeds",
            
"displayName": "Small breeds"
          
}
        
],
        
"id": "dog",
        
"color": "#f6b73c",
        
"type": "list",
        
"displayName": "Dog"
      
}
        
],
    
"enabled": true,
    
"defaultIntentGroupId": "dogs"
  
}
} 

where:
  • accessor - the id of the conversation attribute that comes directly from the external integration and depends on the external system and configurations in it; more details can be found in integration-specific guides
  • defaultIntentGroupId - comes from the Content Mappings configurations
  • id - unique name of the context
  • options - list of context to mapping configurations
  • value - should match a context value set in an external system where Answer Assist is integrated; it is highly dependent on configurations implemented in such an external system
  • intent - comes from Content Mappings
  • color - a color of the context filter field in the hexadecimal format
  • type - a required value that sets the type of the context filter field
  • displayName - will be visible as the dropdown option for agents in the Shelf Answer Assist widget

Figure 39. Adding the needed JSON in the Code Editor of the Advanced tab


                         

Configuring IVR Support via the Answer Assist Context Mappings

Shelf supports integrations with multiple interactive voice response (IVR) systems used to deliver their respective customers the best support via voice services. To make agents’ life more simple and increase satisfaction of their customers with their services, Shelf added the IVR support to Answer Assist. 

It is available out of the box and requires no additional enabling. The only thing needed to be done is to configure the respective mappings similar to standard content mappings. 

Configuring the IVR-specific context mappings is performed as described in the Enabling and Configuring the Context-Mapped Suggestions section above with that difference that the intent value explained in Table 1 above can be used as a respective IVR code the agent needs to select in their platform to get the mapped suggestion(s).

Figure 40. Creating an IVR-specific context mapping in Shelf

For improving average handling time of customer requests, Shelf has implemented the following approach: context-mapped suggestions manually configured by your Admins to be triggered by some particular contact reasons (intent values) open in the Gem tab automatically if: 
  • There is only one Gem offered to you in suggestions
  • No suggestions have been triggered manually in the Answer Assist widget
  • No other suggestions have been displayed before that Gem. 

It should be noted that you or other users are automatically redirected to the Gem tab from any tab set to open by default at the Answer Assist widget launch. 

In this light, Shelf recommends setting the Suggestions tab as the default opening tab for the Answer Assist widget for IVR-specific purposes. It will allow you to avoid switching between tabs and, therefore, deliver accurate and relevant answers to your customers faster.

Figure 41. Viewing the single suggestion in the Gem tab


                           

Configuring Multifield Support in the Answer Assist Context-Mapped Suggestions 

In some cases, customers may have multicomponent questions, for example, a customer asks which small breed dogs are pet dogs. If you go the usual method of the content mapping and context-mapped suggestion configuration, you need to set up several contexts and all their intents and so on. However, Shelf has implemented the multifield support capability for the Answer Assist context-mapped suggestions, and you can set up this capability using the JSON code in the Advanced tab of the Answer Assist Configurator in Shelf. 

This is how the supported multiple fields look in the Suggestions tab of the Shelf Assist widget. They help you filter content more precisely thus allowing you to get access to the needed content faster.

The main idea of this functionality is that you map intent not to some specific Gem URL or search URL with the indicated specific filter values, but to search URL that could contain a combination of one object (where a tag or category of the Gem or a Search term, or even a combination thereof is the filter criteria) and objects with specific values. This allows you or your agents to find the available content depending on the value you or the agents select for this object.

Figure 42. Viewing the Suggestions tab with the enabled feature of the multifield support for context-mapped suggestions

Now, let’s see how you can enable and configure the multifield support for context-mapped suggestions in the Shelf Answer Assist.

To configure this feature, go to the Advanced tab of the Answer Assist Configurator in the Shelf Admin Panel. 
In the Code Editor, add the respective code strings to your JSON responsible for the configuration of the multiple search and filter fields for the context-mapped suggestions.

Below is the example of such code that relates to the search and filter dropdown fields shown in Figure 36.

"contextSuggestions": {
   
"contextTypes": [
     
{
       
"accessor": "dogBreedStr",
       
"options": [
         
{
           
"value": "Large breeds",
           
"displayName": "Large breeds"
         
},
         
{
           
"value": "Medium breeds",
           
"displayName": "Medium breeds"
         
},
         
{
           
"value": "Small breeds",
           
"displayName": "Small breeds"
         
}
       
],
       
"id": "dog",
       
"color": "#f6b73c",
       
"type": "list",
       
"displayName": "Dog Breeds"
     
},
     
{
       
"options": [
         
{
           
"value": "hunting dog",
           
"displayName": "Hunting dogs"
         
},
         
{
           
"value": "mountain dog",
           
"displayName": "Mountain dogs"
         
},
         
{
           
"value": "soviet dog",
           
"displayName": "Soviet dogs"
         
},
         
{
           
"value": "pet dog",
           
"displayName": "Pet dogs"
         
},
         
{
           
"value": "working dog",
           
"displayName": "Working dogs"
         
}
       
],
       
"id": "dog type",
       
"color": "#f6b73c",
       
"type": "list",
       
"displayName": "Dog Type"
     
},
     
{
       
"accessor": "dogBreedStr",
       
"options": [
         
{
           
"value": "Large Dogs",
           
"displayName": "Large dogs"
         
},
         
{
           
"value": "Working Dogs",
           
"displayName": "Working dogs"
         
}
       
],
       
"id": "Size",
       
"color": "#f6b73c",
       
"type": "list",
       
"displayName": "Dog Size"
     
}
   
],
   
"combinations": [
     
{
       
"conditions": [
         
{
           
"context": "dog",
           
"templateVariableName": "breed",
           
"value": "*"
         
},
         
{
           
"context": "dog type",
           
"templateVariableName": "dog type",
           
"value": "*"
         
},
         
{
           
"context": "Size",
           
"templateVariableName": "dog size",
           
"value": "*"
         
}
       
],
       
"intentGroupId": "dogs",
       
"intent": "Breeds"
     
}
   
],
   
"enabled": true,
   
"defaultIntentGroupId": "dogs"

Let’s see what in this code is responsible for multifield support for the context-mapped suggestions. 

"combinations": [
     
{
       
"conditions": [
         
{
           
"context": "dog",
           
"templateVariableName": "breed",
           
"value": "*"
         
},
         
{
           
"context": "dog type",
           
"templateVariableName": "dog type",
           
"value": "*"
         
},
         
{
           
"context": "Size",
           
"templateVariableName": "dog size",
           
"value": "*"
         
}
       
],
       
"intentGroupId": "dogs",
       
"intent": "Breeds"
     
}
   
],
   
"enabled": true,
   
"defaultIntentGroupId": "dogs"


where we are interested in the following arrays - combinations and conditions. 

The first array - combinations - opens that part of the JSON code which tells that there will be a combination of conditions under which the content is to be searched and filtered. 

The second array - conditions - in its turn opens a set of conditions to be configured for the relevant search and filter actions to take place.

It works together with the context configured during the content mapping stage. The code above tells that the single configured intent Breeds (with dogs as defaultIntentGroupId) will be used to search and filter the content for suggestions being offered to the agent IF the three conditions are satisfied - the contexts set to dog, dog type, and Size have any values (“any value” in our case is represented by the asterisk symbol).

"conditions": [
         
{
           
"context": "dog",
           
"templateVariableName": "breed",
           
"value": "*"
         
},
         
{
           
"context": "dog type",
           
"templateVariableName": "dog type",
           
"value": "*"
         
},
         
{
           
"context": "Size",
           
"templateVariableName": "dog size",
           
"value": "*"

If we want to search and filter by some exact value of the context, we need to replace the asterisk with the desired value.

It needs to be noted also that this functionality - multifield support for context-mapped suggestions - works in a combination with the content mapping procedure. And we can see it in this URL used as mappingValue at the content mapping stage - https://yourdomain.shelf.io/?cp=1&ctn={{dog%20size}}&it=card&ps=20&sb=CREATED_DATE&sl=en&so=DESC&tg={{breed}}&tg={{dog%20type}}&tgo=and

In this URL, we can see that the search and filter procedure will return to the agent all Gems which have both the {{breed}} tag and {{dog type}} tags. This is set by the parameter tgo=and.

Figure 43. Viewing the content suggested with the multifield support feature and the AND logic

By default, if no such parameter is specifically entered, the OR logic works, meaning that the search and filter procedure will return either {{breed}}-tagged Gems or {{dog type}}-tagged Gems.  Example of the URL with the OR logic: https://yourdomain.shelf.io/?cp=1&ctn={{dog%20size}}&it=card&ps=20&sb=CREATED_DATE&sl=en&so=DESC&tg={{breed}}&tg={{dog%20type}}b.

Figure 44. Viewing the content suggested with the multifield support feature and the OR logic

In the figure above, we see that the suggestions are searched for and provided to you by any filter that matches one of the selected values - Large Breeds, Mountain Dogs, and Small dogs. In our case, the Shelf Answer Assist suggests Gems that have the Large breeds tag only, because the other conditions are disregarded due to the OR logic.

It should be mentioned that the multifield support feature for the context-mapped suggestions in the Shelf Answer Assist operates based on tags and categories assigned to Gems. 

Figure 45. Viewing tags and categories assigned to a Gem in Shelf

The tags and categories are used as values of the variables. If at least one of the tags or categories used as variables is not assigned to the Gem and the AND logic operator (tgo=and / cto=and)  the Gem, the search does show such Gem in the suggestions in Answer Assist. 

As regards the categories assigned to Gems in Shelf, there are some specifics. First of all, it should be mentioned that Shelf internally uses CategoryId to identify and search for Gems within the Shelf platform.
 
However, external systems may not have such a notion as CategoryID and can be used for context-mapped suggestions only via category names. 

To counter this discrepancy, Shelf has introduced and implemented the category name filtering capability for context mappings. To make use of it, you need to add the ctn={category name} filter into the content mapping URL. In our exemplary content mapping URL, it is here: https://yourdomain.shelf.io/?cp=1&ctn={{dog%20size}}.

If a needed category has subcategories, you can use their names as variables within their parent category, by adding the following parameter to your content mapping URL: ctn={parent category name}.{variable}.

If you try to use a category name that does not exist, in order to get some content-mapped suggestions in Answer Assist, no results are expected to be returned and no Gems are suggested.

If you use a valid subcategory name as a variable for the content-mapped suggestions in Answer Assist but at the same time use a parent category name that does not exist, no results are expected to be returned and no Gems are suggested.

If you use a valid subcategory name as a variable for the content-mapped suggestions in Answer Assist but at the same time do not use its parent category name as a filter in the content mapping URL, the suggestions will contain Gems with the identical category names.

Note about the mapping filtering logic:

If tgo=and is used in a content mapping, only the matching Gems that contain all of the assigned tags are returned in the Answer Assist suggestions.

If tgo=or is used in a content mapping, all the matching Gems that contain at least one of the assigned and indicated tags are returned in the Answer Assist suggestions.

If cto=and is used in a content mapping, only the matching Gems that contain all of the assigned categories are returned in the Answer Assist suggestions.

If cto=or is used in a content mapping, all the matching Gems that contain at least one of the assigned and indicated categories are returned in the Answer Assist suggestions.

If you indicate neither tgo=and/cto=and nor tgo=or/cto=or, the logic for searching and filtering the content for the Answer Assist suggestions falls back to  tgo=or/cto=or.


                         

Configuring Context Filter Fields in Answer Assist

By default, once enabled, the context suggestion filters are displayed on top in the Answer Assist widget as shown in the figure below.


Figure 46. Viewing the default context suggestion filter fields in Answer Assist

Sometimes, you may want to hide the context suggestion filters from the Shelf Answer Assist’s user interface. Or you may need to restrict access to these filter fields for your agents and not show them at all. It may be required by your organization’s business needs or alike. 
Shelf has implemented such a capability for Answer Assist, and you can easily configure how or whether to display these filter fields in the Suggestions tab.
This capability is configured by a single string in your JSON configuration in the Advanced tab of the Answer Assist Configurator in the Shelf Admin Panel.


This string is as follows:

"contextSuggestions": {
      
"visibility": "hidden" | "menu" | "top" 
    
}

where the visibility parameter can acquire one of the three permitted values: hidden, when you want to completely hide the fields from the agents, menu, when you want to take the fields away from the facade of the Suggestions tab and hide it in the menu, and top, which is enabled by default, when the filter fields are always on top in the Answer Assist widget (see Figure 40 above). 

Let see how the Answer Assist widget is to look when the visibility is set to hidden (Figure 47) and menu (Figure 48).

Figure 47. Viewing the Suggestions tab in Answer Assist with the filter fields moved to the menu
Figure 48. Viewing the Suggestions tab in Answer Assist with the completely hidden filter fields

In case the visibility is set to menu, you need to click the menu icon in the Suggestions tab to reveal and use the available filter fields. 

If the visibility is set to hidden, no context filter dropdown fields are displayed and the suggestions are changed only when the new context is auto-detected. 


Hiding Individual Content Mapping Fields in Answer Assist User Interface

In addition to hiding ALL filter fields in Answer Assist’s Suggestions tab, you can also hide individual content mapping fields there. 


This can be done by adding the following code to your JSON in the Advanced tab of your Answer Assist configuration. Note that this parameter needs to be added to the specific content mapping field you want to hide.

"visibility""hidden"


In the general JSON scheme, this parameter is placed together with the content mapping filter field it is supposed to hide from the Suggestions tab.


"contextSuggestions": {

    "contextTypes": [

      {

        "color": "#4682B4",

        "displayName": "Colors",

        "visibility": "hidden",


For example, we have originally three (3) content mapping filter fields displayed in the Suggestions tab in the Answer Assist widget - Cars, Engines, Colors. And we want to hide the Colors field because the dealership now has cars of only one color and, respectively, agents cannot refer to suggested documents where other colors are mentioned. 
Figure 49. Viewing Answer Assist’s Suggestions tab with all content mapping filter fields enabled
So, we go to the Advanced tab of the Answer Assist configuration and add the “visibility”: “hidden” to that content mapping filter field we want to hide, that is to the Colors field.

Figure 50. Hiding the Colors content mapping filter field via JSON


Once we save changes, we can go to the Answer Assist widget and check whether the Colors filter field is or is not displayed.

Figure 51.  Verifying the Colors content mapping filter field is hidden in AA widget


                               

Configuring Suggestion Labels

In some cases, you may also want to show or hide the context labels for suggestions displayed to your agents in Answer Assist. By default, the display of such labels is activated in Answer Assist.

Figure 52. Viewing the context labels for the suggestion in Answer Assist

If you need to hide these labels, you need to change the respective value in one of the parameters in JSON in the Advanced tab of your Answer Assist configuration in Shelf.

To do so, change the boolean value for the displayContextValue parameter to false. This parameter may have only two values: true for the enabled display of context labels and false if they are disabled.

Figure 53. Disabling the display of context labels for suggestions in Answer Assist

Once you have changed the JSON accordingly, click SAVE to confirm your changes. 

Figure 54. Viewing the suggestions in Answer Assist without context labels


                                

Enabling and Configuring Search Suggestions for Answer Assist 

In the light of the anticipated decision to abandon AI suggestions in Answer Assist, Shelf has introduced a new feature aimed to enhance your search experience by providing tailored search suggestions based on the context of your conversations with customers. This feature is called Search Suggestions and aims to help you quickly discover relevant information without the need to manually input search terms. 

The feature operates from within the Search tab of the Answer Assist widget. To start it, hover over the search box.

Figure 55. Viewing Search Suggestions in Shelf Answer Assist widget in Kustomer

What the Search Suggestion feature does is that it reads your conversations with customers and adapts to changes in the context of such conversations. If the context of a conversation changes, the feature analyzes it and adapts to such a change, quickly providing you with the updated suggestions of relevant content to help you answer customer questions in the most precise and relevant manner. 

For configuring and using the Search Suggestions feature with the Shelf Answer Assist, it needs to be enabled for your account by Shelf. Contact your organization’s administrator(s) to find out whether your account is eligible for it. To learn more about the Search Suggestions feature, feel free to contact Shelf at support@shelf.io

If the feature is enabled for your account on the Shelf side, the next step is to enable it for your Answer Assist configuration. 

To do so, perform the following procedure:

1. Go to the Admin Panel, open the Answer Assist Configurator menu, find and open your Answer Assist configuration.

Figure 56. Accessing Answer Assist configuration in Shelf Admin Panel 

2. Once on your Answer Assist configuration page, open the Advanced tab on top.

3. In the Advanced tab, add the following strings to the settings JSON:

"gptSearchSuggestions": {

    "enabled": true

  }

Figure 57. Enabling Search Suggestions for your Answer Assist configuration

Once done, click Save to finalize the setup. Now the Search Suggestions feature is enabled for your Answer Assist configuration and is ready for use.


                                 

Enabling and Configuring Interaction History

The Interaction History feature in Answer Assist makes it possible for you to handle and analyze transferred conversations by sharing interaction’s context. 

This feature allows recording all Gems viewed during your interaction(s) with customers in the Answer Assist’s Gem tab to the new Interaction History tab. The interactions are recorded in a chronological order, with the most recent ones being on top. 

The Interaction History feature works with all types of Gems, including Decision Trees. For DT Gems, the tab even shows what step(s) has(have) been already taken. 

The feature is accessible from the identically named tab. 

Figure 58. Accessing the Interaction History feature in AA 

However, to start using the Interaction History feature to suit your and your colleagues’ needs, it has to be enabled for your Answer Assist configuration first. To do so, follow the below steps.

1. Go to the Admin Panel, open the Answer Assist Configurator menu, find and open your Answer Assist configuration.

Figure 59. Accessing Answer Assist configuration in Shelf Admin Panel
2. Once on your Answer Assist configuration page, open the Advanced tab on top.

3. In the Advanced tab, add the following strings to the settings JSON:

"interactionHistory": {

   "enabled": true

 }

Figure 60. Enabling the Interaction History feature and tab for Answer Assist

Once done, click Save to finalize the setup. Now the Interaction History feature is enabled for your Answer Assist configuration and is ready for use.


 

Previewing the Answer Assist Widget in Shelf

Shelf sports a great feature that enables you to see the results of the settings you configure for the Answer Assist widget. It resides in the Preview tab of the Answer Assist Configurator.

Figure 61. Using the Answer Assist Preview feature in the Answer Assist Configurator

Using the Preview feature, you can make changes to the Answer Assist configuration in the Shelf Admin Panel, including the advanced settings via JSON, and, after saving the changes, see how these impact the Answer Assist widget. The feature shows all the enabled Answer Assist widget tabs, suggestions, favorites, announcements, and allows for filtering and evaluating the suggestions, searching for the content, adding or removing it to or from the favorites, responding to announcements and so on.

Enabling the Public Share Feature for Answer Assist

The Public Share feature used to share Gems publicly needs to be enabled for your account on the Shelf side. 

Once it has been enabled for your account by the Shelf team, you need to enable it in the Customize menu in the Shelf Admin Panel. Performing it, you can also set the Gem link expiration period. The minimum period is 1 day, and the maximum is 365 days.

Figure 62. Enabling the Public Share feature in the Admin Panel

The final step is to grant the relevant permissions for the relevant users. This is done in the User Groups menu of the Shelf Admin Panel. Select the needed user group, navigate to the Library Permissions tab and select the needed user role(s) from the dropdown menu, and then select the Share publicly checkbox.

Figure 63. Granting the Public sharing permissions for users

When all the above configurations are complete, you or the users you have granted the respective permissions can share the Gems to the public, even to those recipients who have no Shelf subscription or account. The Gem is shared by means of a publicly accessible web link. 

Multiple Configurations Support in Answer Assist

The Shelf Answer Assist product supports multiple integration configurations, in order to ensure that all needed sources of content can be properly set up for your Answer Assist. The maximum number of supported configurations is 20, each of which can be configured for a separate content input. 

You can create, configure, and manage configurations immediately after accessing the Answer Assist Configurator menu from the Shelf Admin Panel.

Figure 64. Viewing the list of available Answer Assist configurations in Shelf

Each of your Answer Assist configurations can acquire different general settings - suggestion sources, features enabled/disabled - and advanced settings - widget tabs orders, default opening tab, Gem linkage, etc.

You can access and update or delete any of the available configurations. Click the needed one or click the gear icon next to it, in order to access the configuration’s editing mode, in which you can delete the selected configuration or introduce the needed changes to it. See the Creating and Deleting the Answer Assist Configuration section for more details.



3.2. Shelf Answer Assist Features Overview

Now, when the necessary configurations have been implemented on the Shelf side, we can walk you through the features hosted by the Answer Assist product.
We will guide you through the standard scenario of the Answer Assist usage that can be as follows:
  1. During an interaction with a customer, you start the Answer Assist widget and get the relevant Suggestions (both AI-matched and context-mapped suggestions)
  2. If no Suggestions are provided by Answer Assist, you use its Search feature to look up for the needed help articles (Gems)
  3. Before running a manual search to find content, you may use the Search Suggestions feature directly in the Search tab of the Answer Assist widget and receive the list of three most suitable suggestions based on your interaction with the customer
  4. Upon finding the suitable article, you view it in the Gem tab of the Shelf Answer Assist widget and can also link the article to the interaction from tooltips and open it in Shelf. You can also open any found article (Gem) not in Answer Assist’s Gem tab but in a new tab of your browser for a quick preview via the combination of Command (MacOS) or Control button on your keyboard and the left button of your mouse.
  5. After reviewing the article, you provide your feedback on it
  6. If you find the content in the article meaningful and helpful, you can share it with your customers or colleagues using the Public Share feature
  7. In addition to sharing the article, you can add it to Favorites for future reference; you can also remove other previously added article(s) from favorites
  8. Check the Announcements tab for any updates in the content sharing policy or the customer support regulations etc.
  9. Work with the Links tab in Answer Assist: check whether you have linked all the needed articles to the relevant interaction, unlink no longer needed articles etc.


3.2.1. Answer Assist Features Usage

 

Receiving and Working with Real-Time Suggestions

One of the Shelf Answer Assist’s features is to provide real-time content suggestions to your agents to allow them to address your customers’ questions in a quicker and more efficient fashion. The Real-Time Suggestions feature was developed to substitute the now-deprecated AI Suggestions feature and ensures a higher level of suggesting accuracy and speed.

Let us see how you, as an agent, or your organization’s agents can leverage this feature to pursue your business goals.

The real-time suggestions are generated and delivered to agents during both chat and voice conversations with customers. 


Example 1. Generating and receiving real-time suggestions in Answer Assist chat on Kustomer

Figure 65. Viewing real-time suggestions being generated during chat conversation on Kustomer
Figure 66. Viewing generated real-time suggestions in Answer Assist widget on Kustomer 

Example 2. Generating and receiving real-time suggestions in voice conversation on Genesys PureCloud
Figure 67. Viewing real-time suggestions being generated during voice conversation on Genesys

Figure 68. Viewing generated real-time suggestions in Answer Assist widget on Genesys


Speaking about real-time suggestions in Shelf Answer Assist during voice conversations on the Genesys PureCloud platform - it is a good example of applying dynamic filters to such suggestions.


These dynamic filters are defined by the following parameters and their values, which need to be added to the JSON code in the Advanced tab of Shelf Answer Assist Configurator menu, as described in the Enabling and Configuring Real-Time Suggestions section.


{

    "filtersMapping": {

          "collectionIds": "Genesys.collectionids",

       "categories": "Genesys.categories",

       "tags": "Genesys.tags

    }

}


These parameters set mappings between Shelf and Genesys, defining what Shelf parameter corresponds to what Genesys parameter. 

When you access your Genesys PureCloud instance and create the inbound voice call flow, you define these parameters there as well. See the following figure for details. 

Figure 69. Configuring CollectionIds, categories, tags parameters on Genesys


Once these parameters are set and defined, Answer Assist will filter real-time suggestions based on the relevant content.


Regardless of the flow—chat or voice conversation—of receiving real-time suggestions in Shelf Answer Assist, you can work with them as follows:

  • Initially view 3 (three) most suitable suggestions (1) and the customer question they are generated for (2)

  • View which of the displayed suggestions are new and which are not (3)

  • Access the Gem tab for receiving more detailed information on the issue (4)

  • Expand (5) the suggestions card and view up to 5 (five) suggestions in total (6)

Collapse the suggestions back to 3 (three) options (7)
Figure 70. Actions with real-time suggestions in Shelf Answer Assist widget


      

Working with the Context-Mapped Suggestions in Answer Assist

Once all the settings described in the Enabling and Configuring the Context-Mapped Suggestions section have been made, you can check how the Answer Assist delivers context-mapped suggestions via the widget.

Figure 71. Checking the context-mapped suggestions in Answer Assist in Kustomer

The figure above demonstrates that the Answer Assist has absorbed and processed the customer’s question during the interaction and provided the agent with the matching content suggestions in the widget. 

If needed, you can select a suitable context option from the context filter’s list. It will streamline the suggestions offered.

Figure 72. Using the context filtering options in Answer Assist

If the multifield support feature for context-mapped suggestions has been enabled and configured for your Answer Assist, you can filter the content even deeper, using combinations of available content filters.

Figure 73. Viewing the filtered context-mapped content suggestions in Answer Assist


       

Using Search Suggestions to Obtain Relevant Content Recommendations

If the preconfigured context-mapped suggestions described in the section above are not sufficient for your needs or do not provide sufficient accuracy in content findings, or there occur unexpected changes of context in your conversations with customers, you can improve both your and your customer’s experience and satisfaction by using the Shelf Answer Assist’s Search Suggestions feature. 

How it works
During a conversation with your customer the system reads its messages and dynamically generates up to three (3) search suggestions based on the conversation context. What you need to do to make use of the Search Suggestions feature is to open the Search tab (1)  in your Shelf Answer Assist widget and click anywhere in the Search field (2). Up to three most suitable search suggestions will be displayed underneath, making it much faster and easier for you to look for the needed information.

Figure 74. Viewing system-generated search suggestions in Shelf Answer Assist

After receiving the search suggestions, you just select the most suitable one by clicking it and get to the list of content items that may contain the information needed for your customer. 

Figure 75. Viewing content items relevant to the selected search suggestion 

After this step you just follow your usual procedure to answer your customer’s question(s). 

However, the most useful capability of the Search Suggestions feature is its ability to adapt to changes in the context of your conversations with customers. For example, if the initial question of your customer was about dogs but then they suddenly ask you how or where they can purchase a car, you no longer need to run a separate search - Answer Assist will provide you with the relevant search suggestions automatically once you click on the Search field.

Figure 76. Viewing search suggestions for a changed conversation context in Shelf Answer Assist

Note:
Please remember that the Search Suggestion feature is powered by chatGPT and needs to be treated accordingly, meaning that in rare cases it can provide suggestions not fully matching your needs.


           

Looking for Content using the Search Feature and Search Tab

If the suggestions offered by Shelf’s Answer Automation feature does not work for you, you can always look for the needed information in your Shelf or external (if the Federated Search feature is enabled and configured) libraries using the Answer Assist Search tab.

Figure 77. Viewing the Search tab in Answer Assist

You can use either a free search by entering your request in the search bar in the upper part of the tab window, or use the advanced search options that appear once you click the hamburger menu button in the uttermost right part of the search bar.

Figure 78. Using advanced search options in Answer Assist
Here are the details of the advanced search settings:
  1. Library
    You can limit the search by some specific library only. For this purpose, click the field and select the needed library from the dropdown list. If the Federated Search feature is enabled for your Answer Assist instance, you can select multiple libraries.

    Figure 79. Selecting the needed library(s) to search in

  2. Content Type
    You can select which type of content you want to look for. For this purpose, click the field and select the needed option(s) from the dropdown list.

    Figure 80. Selecting the type of the searched content

  3. File Format
    You can also select what type the content you look for needs to be. To do so, click the field and select the checkbox(es) next to the needed file type option(s).

    Figure 81. Selecting the file format for your searched content

  4. Language
    If, like in the case of the similar filter in the Suggestionstab, multilanguage support is enabled for Answer Assist on the Shelf end, you may look for content in some certain language only. For this purpose, click the field and select the needed option from the dropdown list that appears.

    Figure 82. Selecting the language of the search content

  5. Source
    You can indicate the source of the content you are looking for. It can be either Shelf or some third-party or external platform, or any. Click the field and select the needed option from the dropdown list.
    Figure 83. Selecting the content source
  6. Added by
    If you know who has added the content you are looking for, you can indicate this user in your search to narrow it. For this purpose, click the menu icon in this field. In the field that appears next, look for the needed user(s) in the search bar. The search-as-you-type feature is supported and yields the suggested user names under the search bar. Once the needed user(s) has/have been found, select them and click the SEARCH button to trigger the content search or the Backicon to go back to the advanced search settings.

    Figure 84. Selecting users who added the content to refine the search

    You can select as many as 20 users to your search, using comma as a separator. You can also see how many content items have been added by each of the users. The count is shown next to the user’s name. Use the up and down arrows to sort the users in alphabetic order from A to Z or from Z to A, or sort them by the number of gems they have added, in ascending or descending order.
  7. Rating
    You may refine your search by indicating the rating you want the content to be. For this purpose, click the field and select the needed option from the dropdown list.

    Figure 85. Selecting rating of the search content

  8. Category
    It is also possible to look for content of a certain category only. The category in this case is one of the categories assigned to Gems on the Shelf side. To select the needed category(ies), click the menu icon and, in the window that appears next, search for the needed one(s). The search-as-you-type feature is supported, and the results are displayed below the search bar. All the available categories are also displayed under the search bar. Select the needed category(ies) and click the SEARCH button to trigger the content search or the Backicon to go back to the advanced search settings.

    Figure 86. Searching for content from the specified categories

  9. Tag 
    Finally, you can look for content that has only some specific tag(s). The tag in this case is one of the tags assigned to Gems on the Shelf side. To select the needed tag(s), click the menu icon and, in the window that appears next, search for the needed one(s). The search-as-you-type feature is supported, and the results are displayed below the search bar. All the available tags are also displayed under the search bar. Select the needed tag(s) and click the SEARCH button to trigger the content search or the Backicon to go back to the advanced search settings.

    Figure 87. Searching for content with the specified tags only

You can also sort the tags by name from A to Z or by the number of gems having them, in ascending or descending order, using the up and down arrow icons.

Once you have configured your search settings, click the SEARCH button. The search results will be shown in the window.

Figure 88. Viewing the content search results

You can see the total number of the content items and how many of them are displayed on this page. 
You also can refresh the search results by clicking the Refresh icon (1). 
You can also clear the search results by clicking the CLEAR ALL link button (4). 
You can view any of the found content items by clicking it (3) and opening it in the Gem tab. 
Finally, you can open these search results directly in Shelf by clicking the respective icon (2). 
 
Figure 89. Viewing the AA search results in Shelf



Working with Content in the Gem Tab

The Gem tab is where the content is displayed. It can be the content from various locations within Answer Assist: the Search tab, the Suggestions tab, the Favorites tab, the Announcements tab, and the Links tab also. Once you have clicked a Gem card or header, it opens in the Gem tab.

Figure 90. Viewing the Gem tab in Answer Assist

The Gem tab has the following structure:
  • Actions panel
  • Breadcrumbs 
  • Gem title, description, and content block
  • Connections
  • Gem details (Gem metadata)
  • Feedback tools (if the Feedback feature is enabled for your Answer Assist configuration)

Let us see now what you can do with the Gem in the Gem tab, using its structure and available tools.

Using the Actions Panel

The Actions panel is a block of clickable icon buttons - Link/Unlink, Add to Favorites/Unfavorite, Open Gem in Shelf, and Share the Gem publicly.

Figure 91. Viewing the Actions Panel in the Gem tab in Answer Assist


       

Linking/Unlinking Gems to/from Interactions 

If the Gem Linkage feature is enabled, an additional button appears on top of the Gem tab in Answer Assist - Link/Unlink. You can use it to link the Gem to the needed interaction with a customer or unlink it from it. 

There are two ways to connect (link) a Gem to an interaction. 
The first way is to link the Gem to your current interaction with the customer. To do so, when in the Gem tab, click the Link/Unlink button in the Actions Panel.

Figure 92. Linking a Gem to an interaction in the Gem tab

Once you have linked the needed Gem to the specific interaction with the customer, you can open the Links tab in Answer Assist to verify the Gem has been linked.

Figure 93. Verifying the Gem linking to the interaction in Answer Assist

The second way of linking a Gem to an interaction is to use the Links tab. In this case, you can link ANY Gem to ANY interaction, even the past one. 
To do so, follow the steps below.
  1. Enter the needed Interaction ID in the respective field. The ADD button becomes enabled. 
  2. Enter an ID or URL address of the needed Gem in the respective field that appears on the screen and click the LINKbutton. You can find both the Gem URL and ID in the address bar once you are viewing the needed Gem on Shelf. See the figure below for details.

    Figure 94. Finding Gem URL and/or ID on Shelf

  3. Once you have found and entered either the Gem URL or ID, click the LINKbutton.

    Figure 95. Linking a Gem to an interaction in Answer Assist 


Once you are done, the Gem is linked to the needed interaction and will be displayed in the Links tab when you enter the respective Interaction ID next time.
Note that the maximum number of Gems you can add to one interaction is 100.

In order to unlink the Gem from the interaction, go to the Gem tab and click the Link/Unlink button again.

Figure 96. Unlinking the Gem from the Interaction

You can also unlink the Gem from the interaction directly from the Links tab. Find the needed Gem in the list of Gems linked to the respective interaction and perform the below steps.
  1. Hover over this Gem to reveal the Unlink icon next to it.
  2. Click the Unlink icon to remove the Gem from the list of Gems linked to the interaction. When the Gem becomes unlinked, you will see the respective notification.
  3. You can also revert the Gem unlinking by clicking the UNDO button in the above notification. Note that you can revert the Gem unlinking only during 5 seconds after you have unlinked the gem. This is how long the Gem is unlinked notification remains on the screen before disappearing.

Figure 97. Unlinking a Gem from an Interaction in Answer Assist


                 

Adding/Removing Gems to/from Favorites 

You can also add the useful or helpful Gems to Favorites or remove them from there when they are no longer needed. You can do it if the Favorites tab has been enabled for your Answer Assist configuration in the Answer Assist Configurator in the Shelf Admin Panel.

To add the Gem to Favorites, click the respective - heart - icon in the Actions panel in the Gem tab page.

Figure 98. Adding the Gem to Favorites in Answer Assist

Check if the Gem has been added, by opening the Favorites tab.

Figure 99. Verifying the Gem addition to Favorites

In order to remove the Gem from Favorites, open the needed favorited Gem and click the heart icon again. The Gem becomes unfavorited.


                   

Opening Gems in Shelf 

Next thing you can do via the Actions panel in the Gem tab is to open the Gem in Shelf. 
To do so, click the respective icon. 

Figure 100. Opening the Gem in Shelf

A new tab/window will open in your browser where the selected Gem is displayed in Shelf. 


Figure 101. Viewing the Gem in Shelf


                   

Sharing Gems to Public

The fourth option in the Actions panel is only operational if the respective feature - Public Share - is enabled. 

Once the Public Share feature is enabled, the Actions panel acquires the fourth clickable icon button that allows you to share the link to the Gem to the needed recipients.

Figure 102. Sharing the Gem publicly

Once you confirm your desire to share the Gem publicly by clicking the SHARE button, the Gem link is copied to the clipboard and you can share it with anyone - no Shelf subscription is required to view the Gem.

Figure 103. Viewing the publicly shared Gem

Moreover, if a Gem you have received as a suggestion or found via Answer Assist’s Search feature contains another linked Gem in it, you can share this linked Gem publicly without opening it either in Answer Assist’s Gem tab or in Shelf.  All you need to do is to hover the pointer over the linked Gem and then click the Share Gem as a public link button in the Gem preview tooltip.

Figure 104. Publicly sharing the Gem linked to the currently opened Gem

Using Breadcrumbs

Breadcrumbs in the Gem tab identify the location of the Gem, that is, show in which library, folder, or subfolder it is stored. 

Figure 105. Viewing the breadcrumbs in the Gem tab in Answer Assist

Using the Gem Block

                      

Viewing Gems and their Content

In the Gem block of Answer Assist’s Gem tab, you can see the title (name) of the Gem, its description, and read its content.

Figure 106. Viewing the Gem and its content in Answer Assist


                        

Copying/Sending Gem’s Content

Additionally to viewing the Gem title, description and reading its content, you can also copy and send the content - text or image(s) - of the Gem. 

If you select any text or image in the article in the Gem tab, you will be able to copy the selection for further use (by clicking the Copy button), or send the selection directly to your chat with the customer (by clicking the Send button).

Figure 107. Copying or sending the Gem content


                          

Providing Feedback on Gems

The Feedback tools you can see in the Gem block of Answer Assist’s Gem tab allow you to leave your feedback on the Gem - whether it is helpful or not for the resolution of the customer’s issue. Click the needed emoji to rate the Gem accordingly.

Figure 108. Leaving the feedback on the Gem in Answer Assist

Though the Feedback feature is available in Answer Assist for suggestions by default, it needs to be switched on separately in the Answer Assist Configurator in the Shelf Admin Panel. This implementation has been selected in order to have seamless integration with the Shelf Feedback Manager feature.

Once you provide your feedback on the Gem (upvote or downvote it), it becomes recorded and is further used to improve the quality of suggestions provided in Answer Assist. 

More details on the Shelf Feedback Manager can be found in the respective feature-specific guide.

Using the Connections in the Gem Tab

Immediately after the Gem block - Gem’s content, that is text, graphics, links etc. - you can see the Connections section in the Gem tab. It is only visible if the Gem you are currently viewing has any connected Gems. 

Figure 109. Viewing Gem’s connections in Shelf

Connections are displayed as an indefinite scrollable list right under the Gem’s content and feedback tools sections. The Connections block is collapsible and you can expand it if needed or collapse it if not needed. The number of connections in this block is unlimited - all Gems connected to the currently viewed Gem are displayed in this block. 

Functionality of the Connections block is similar to the Suggestions - you see headers and labels of the connected Gems and can click them to view. Once you click any Gem in the Connections block, it opens in the same Gem tab as the Gem you are currently viewing.

Figure 110. Using the Connections block in Answer Assist’s Gem tab


                            

Using the Gem Details in the Gem Tab

In the Gem details section, you can view various information about the Gem, such as: author, creation date, change date, language, category and tag assigned to the Gem, as well as language and source of the Gem. Basically, it is the metadata summary of the Gem.

Figure 111. Viewing the Gem details in the Gem tab of Answer Assist


                              

Using the Personal Settings Tab in Answer Assist

Personal Settings is a tab in the Answer Assist widget which you can use to log out of the Shelf Answer Assist widget. To do so, click the respective SIGN OUT button.

Figure 112. Viewing the Personal Settings tab and signing out of Answer Assist


                               

Using the Announcements Feature in Answer Assist 

The Announcements feature in Shelf and in Answer Assist you to receive notifications about important content changes or updates, or notifications about any other events or publications associated with your organization. 

This feature is enabled and configured on the Shelf side but your organization’s admins need to switch it on for your Answer Assist configuration in the Shelf Admin Panel.

Announcements are used to track how your organization’s agents comply with the established content publishing and use procedures - content managers of your organization can see who exactly has received the notification about some content changes or updates and understand whether this agent has acknowledged its receipt.

At the same time agents within your organization can view such announcements and confirm reading and understanding some important information they contain, directly in the Shelf Answer Assist widget.

Figure 113. Viewing the Announcements tab contents

In this tab, you see the announcements sent to you by your content manager(s).

Some announcements may require you to confirm that you have read them by selecting the checkbox next to the respective text. 

Figure 114. Viewing the announcement that require confirmation

These announcements will have the dedicated Confirmation needed label as shown in Figure 97 above. If an announcement is new and unseen yet, it will have the New label. 

Some may be just informative and require no confirmation. The other announcements may have attachments - Gems - that contain some additional important information for you. You can open such Gems by simply clicking them. 

More details about the Announcements feature can be found in the dedicated Announcements Feature Overview Guide.


                                                               

Using the Interaction History Feature in Answer Assist

As mentioned in the Enabling and Configuring section above, the Interaction History feature records your interactions with customers in terms of the content viewed and distributed so that you or your colleagues are able later, when reopening the specific interaction, to view what content has been viewed, when it has been viewed, and who has viewed it during the interaction. 

To access the Interaction History feature, open the tab with the same name.

Figure 115. Opening and viewing the Interaction History tab in the Answer Assist widget

In this tab, you are able to see all Gems you or your colleagues have viewed during an interaction in the Gem tab in Answer Assist. The Gems are listed in a chronological order and sorted by the most recent Gems on top. If you hover the pointer over the info icon, the relevant tooltip is displayed.

Figure 116. Viewing the content sorting tooltip in the Interaction History tab

In the Interaction History tab, you can also see 
  1. location from which a relevant Gem has been opened
  2. date and time when this Gem has been opened from and the time when it happened will be indicated by corresponding labels and tooltips with more detailed explanations
  3. user(s) who opened the Gem(s), as all the Gems in the Interaction History tab are sorted under their viewer names.

Figure 117. Interaction History data explained

You can go to the respective Gem directly from the Interaction History tab by simply clicking it. The Gem is opened in the Gem tab, respectively.

Figure 118. Opening and viewing Gem from the Interaction History tab in the Gem tab

If any of the viewed Gems is a Decision Tree (DT), you can see its Journeys and review what steps have been already taken. The DT Journey steps are displayed in the same manner as regular Gems, that is the most recent being on top. 

Figure 119. Viewing DT Gem with the collapsed steps in the Interaction History tab

Figure 120. Viewing DT Gem with the expanded steps in the Interaction History tab

If you open the desired DT Gem from the Interaction History tab, you are redirected to the step on which this Gem has been last opened during the interaction. This has been implemented to save your time and make it easier to continue your existing conversation with a customer from where it has been paused.

Figure 121. Viewing DT Gem on the last opened step in the Gem tab

If a Gem listed in the Interaction History tab has been deleted in Shelf KMS, it is displayed in the tab as a card with the trash bin icon and the warning text reading “Gem has been deleted”. Also, such a Gem is not clickable, that is you cannot open it in the Gem tab in Answer Assist widget nor in Shelf KMS. However, the Answer Assist widget chips show when and from which tab this deleted Gem has been opened last time.

Figure 122. Viewing the existing Gem in the Interaction History tab 

Figure 123. Viewing the deleted Gem in the Interaction History tab 

If there is a Gem listed in the Interaction History tab which you have no permissions to view, the Gem card shows the lock icon and the text reading “You have no permissions to this gem”. The Gem is not clickable, that is you cannot open it in the Gem tab of the Answer Assist widget. You also cannot see its title in the Interaction History tab.

Figure 124. Display of the Gem with no permissions to view




4. Conclusion

Thank you for reading this End-to-End Guide. We tried to cover as much as possible in terms of the Answer Assist product features and capabilities. We hope that you now are able to enable, configure and use all of it in an efficient manner, fully knowing what each of such features and capabilities is meant for. 


In case you need any additional assistance on Answer Assist or anything related to it, feel free to contact our Support Team at support@shelf.io or from the in-app chat within the Shelf platform.