Deploying a Virtual Agent


A virtual agent assembles all knowledge for a set of Self Help projects, and uses it to communicate with users within a given scope via a dialogue of questions and answers.
         Virtual agent overview - EN.png

The virtual agent is deployed in several phases: 

  • Project design and preparation.
  • Import or create all Self Help projects required.
  • Adapt procedures to the identity and scope of action of the virtual agent in compliance with the style guide.
  • Create the virtual agent and assemble all data.
  • Deploy the virtual agent and ensure continual improvement of the Knowledge Base.

Project design and preparation

Functional scope of the virtual agent

The functional scope describes the scope of action for the virtual agent. It defines the topics that the virtual agent can cover in conversations with users, as well as the knowledge it requires to answer their questions.

  • The functional scope must be clearly delimited and defined for the virtual agent's specific role.
  • It must be considered:
    • In terms of integration, to configure the display of the virtual agent's answers correctly depending on the channel, i.e. Service Apps app, Web page, or Microsoft Teams collaborative platform.
               Open url.png See List of available channels.
       
    • In terms of interfacing with third-party products, to install the connection packages required for the tool and the virtual agent to communicate.
               Open url.png See List of available connection packages.

Examples

A virtual agent can play the following roles:

  • Manage access to a Knowledge Base, e.g. Support database, HR database, etc.
  • Integrate a business process, e.g. create, process or close tickets.
  • Provide automated answers to common customer questions.
  • Assemble knowledge from different tools, e.g. IT database, Salesforce database, etc.
  • Guide visitors to the right product or service.

Best Practice

  • To deploy your virtual agent rapidly, define a limited functional scope during the design phase, focused on minor tasks. Then extend the scope progressively based on user feedback (continual improvement phase).

Ergonomic aspects of the virtual agent

Several elements enable you to create a well-defined identity for the virtual agent to make it easily identifiable for users.

Personality of the virtual agent

Each customer can define a different personality for their virtual agent, with specific characteristics such as:

  • Identity: Assign a name, gender (male or female), visual identity with an avatar, etc.
  • Style of language depending on the target audience: Technical or non-technical vocabulary, informal or formal style of language, verb or noun, sense of humor, etc.
  • Level of interaction: A proactive virtual agent that initiates the conversation, or a passive virtual agent where user questions are expected first.

Best Practice

  • Define your style guide with compliance rules based on the personality you defined for the virtual agent.

Visual identity of virtual agent

The virtual agent is based on the evSH skin. However, several elements can be configured based on the visual identity you assign the virtual agent: 

  • Logo, sub-heading, virtual agent name, color of headings, etc.
  • Dominant colors and alternatives
  • Font and size

Calling the virtual agent

The virtual agent can be called in different ways:

  • A call button on each page, used to initiate a conversation
  • Automatic opening of the conversation panel as soon as the virtual agent's welcome message appears
  • Automatic opening of the conversation panel as soon as a question is asked, to initiate a new conversation

Best Practice

  • If the virtual agent is intended to relieve the load of other channels (e.g. Support website), you should encourage users to use this self-help solution. Ensure that the virtual agent is visible, i.e. more than just a button at the bottom of the page.

Characteristics of the conversation panel

Several elements can be configured for the conversation panel:

  • Font and size of text displayed in the panel
  • Color of the panel header and footer
  • Predictive text
     

Configuration is performed using the Virtual Agent widget if the virtual agent's call button is integrated in a Service Apps app.

Creation of the virtual agent

Structure of the virtual agent

The virtual agent assembles all knowledge for a set of Self Help projects associated with a dedicated Self Help domain, e.g. Social Base project, other projects using third-party product data (Service Manager, Salesforce, etc.), customer projects.

  • The Self Help projects are created / imported via the Studio Edit mode. Then they are accessible via the tree structure.
  • They are arranged in folders and subfolders to facilitate access to procedures and resources.
  • Resources are managed via a dedicated folder in each subfolder, created near the procedures that use them. Shared resources are managed at the root of the Self Help project.

Notes

  • The Social Base project is vital for the virtual agent to work.
  • The virtual agent can be designed rapidly using templates.
    • Each template can import sets of procedures and functionalities specific to a role and a precise scope of action for the virtual agent. 
    • A new Self Help project is automatically created when a template is imported, with its entire tree structure.

Best Practice

  • Create a Self Help project for each third-party product used by your virtual agent. Installation is easy using connection packages.
  • To manage access rights, create distinct Self Help projects to consolidate all procedures that can be accessed by different groups. Next, assemble all these projects in your virtual agent.
  • Only create procedures dedicated to your virtual agent. Avoid grouping them with the procedures for your customer portal.
  • Isolate all knowledge shared by virtual agents in dedicated Self Help projects within the same domain.

example Sharing Social Base and Security database Self Help projects among several virtual agents

         Several virtual agents - Schema - EN.png

User and writer authorizations

Users

  • User authorizations for communicating with the virtual agent (access to the virtual agent's Self Help project procedures) are managed in the group to which the user belongs.
  • User groups can be created using various criteria, e.g. department, location, level of skill, etc.

Writers

Writers are usually subject knowledge experts. They write and publish procedures. 

  • They are authorized to work on the virtual agent's Self Help projects based on their read, write and publish rights.
             Open url.png See Access rights management.
  • They are supervised by the Knowledge Manager.
    • The role of this person is to ensure the uniformity of the virtual agent's Knowledge Base, ensuring that each procedure is written in the same way in compliance with the style guide. 
    • The Knowledge Manager also makes sure that the topics covered by the virtual agent fall within the functional scope.

Adapting procedures

The procedures for Self Help projects for the virtual agent must be adapted to its identity and its scope of action, as well as the customer-specific style guide.

The behavior of procedures can also be adapted:

  • By using dedicated metadata to alter the behavior of the procedure when run by the virtual agent.
  • By using wildcards to make the procedure dynamic.

Notes

  • A large number of common errors are stored in the NLP (natural language processing) engine used by Service Bots, and are automatically corrected by the virtual agent when users enter their questions.

example Correction of typos and spelling mistakes

         NLP engine - Corrections examples.png

Best Practice

  • Define your style guide with compliance rules based on the personality you defined for the virtual agent.
  • Insert informative and non-illustrative images to keep the virtual agent's answers brief.
  • Your virtual agent can guide users after providing an answer, by suggesting other content related to the same topic. In this case, you should make sure that users do not feel blocked in the conversation.
             Open url.png See Content suggestions.
  • To ensure the uniformity of the Knowledge Base for your virtual agent, create procedure templates that are accessible to each writer, e.g. standard step-by-step procedures, escalation procedures, satisfaction surveys, connector call procedures, etc.
    • Group these templates in a dedicated hidden folder at the root of the Self Help project.
    • Name the folder "[999] Templates" so that it is systematically found at the bottom of the tree structure.
    • Prior to the publication of the Self Help project, check that the procedures in this folder cannot be run by users.
  • To help writers comply with the style guide when writing procedures, implement the content validation workflow for all Self Help projects for your virtual agent. In this way, writers and the Knowledge Manager will be required to follow all approval steps before procedures can be published.

Deployment of the virtual agent

Once all of the Self Help project procedures have been created or adapted for the virtual agent, they can be published. The virtual agent can then be deployed and made available to users.

Creation of the virtual agent

Creating the virtual agent consists of assembling all data from Self Help projects, e.g. Social Base project, projects using third-party product data (Service Manager, Salesforce, etc.), customer-specific Self Help projects.

You create the virtual agent in the Administration module, which can be accessed via the Editor Portal. Open url.png See the procedure.

Monitoring and continual improvement

The purpose of this phase is to analyze the use of the virtual agent's Knowledge Base and to enhance its content by adapting procedures for more relevant answers, or by creating new procedures for unexpected questions, etc.

Several tools are available for this purpose.

  • Feedback from end users.
  • Usage statistics for each of the virtual agent’s procedures, accessible via the Properties pane in the Studio Edit mode. They are used to view the number of executions and their average duration.
  • Statistical reports, accessible in the Administration module on the Editor Portal (Analytics module). They display the most frequently used procedures and the most rarely used ones, keywords generating no results, etc.

Best Practice

  • Perform an in-depth analysis of statistics by replaying user journeys to see the actions users took, the questions asked, etc.
  • Encourage users to leave comments (misunderstanding, missing or incorrect content, issues regarding procedures, etc.) in a satisfaction survey at the end of each procedure.

Procedures

How to implement a virtual agent

Step 1: Design and prepare the project.

    Note: Step performed by the administrator.

1. Define the functional scope and characteristics of your virtual agent. 

2. Define its personality (gender, vocabulary used, etc.) and visual identity based on the target audience.

3. Define the style guide for your virtual agent and procedures.

4. Define the display channel to be used for the virtual agent and the call method.

5. Identify third-party products with which the virtual agent must communicate.

Step 2: Install the Social Base package.

Note: The installation of the Social Base package is vital for the virtual agent to work.

1. Download the Social Base package used by the virtual agent.Open url.png See the description.
         Download icon.png Social Base

2. Go to the Studio Edit mode and import the file into the domain of the virtual agent.
A new Self Help Social Base project is created. All procedures are automatically imported and added to the tree structure.

3. Adapt the procedures to the identity of your virtual agent. Open url.png See How to customize the Self Help Social Base project.

Step 3: Install connection packages.

1. Download all connection packages used by your virtual agent.
         Open url.png See List of available connection packages.

2. Go to the Studio Edit mode and import each file into the domain of the virtual agent.
A new Self Help project is created for each connection package. All procedures are automatically imported and added to the tree structure of each project.

4. Configure the access to each new Self Help project. See the relevant documentation.

5. Adapt the procedures to the identity of your virtual agent.

Step 4: Install the Service Bots package.

1. Download the Service Bots package to be integrated into your virtual agent.
         Open url.png See List of available templates.

2. Go to the Studio Edit mode and import the file into the domain of the virtual agent.
A new Service Bots project is created. All procedures are automatically imported and added to the tree structure.

3. Adapt the procedures to the identity of your virtual agent.

Step 5: Create the customer-specific Self Help project.

    Note: Step performed by the administrator.

1. Create the new Self Help project in the domain of the virtual agent.

         Open url.png See How to create a Self Help project.

  • Create the tree structure of folders and subfolders. 
  • Organize resources.
    • Define a folder at the Self Help project root for all shared project resources.
    • Define a subfolder in each procedure folder for all resources used by this procedure.
    • Define the naming convention for each type of resource.
       

    Note: Step performed by writers.

2. Create and write the procedures for your customer-specific Self Help project.

         Open url.png See How to create a Self Help  procedure.

Step 6: Publish the virtual agent procedures.

    Note: Step performed by writers.

1. Publish all procedures for Self Help projects.


Step 7: Deploy the virtual agent.

    Note: Step performed by writers whose profile is authorized to access the Administration module.

1. Open the Administration module in the Studio Home page by selecting Editor Portal > Administration.

2. Select Administration > Service Bots > Agents in the menu.

3. Click Add to create your virtual agent.

4. Enter its name, reference and description.

5. Select its dedicated Self Help domain.

6. Click Add icon.png and Add to select all Self Help projects for the agent, i.e. Social Base project, connection packages, Service Bots package, your customer-specific Self Help project.

7. Click Add.
Your virtual agent is created.

8. Configure the integration mode of the virtual agent depending on the display channel. Open url.png See the procedures.

How to integrate the virtual agent depending on the display channel

Integration in a Service Apps app

1. Open the Service Apps app in the Studio Edit mode.

2. Add a Virtual Agent widget.

3. Configure the virtual agent. Open url.png See the procedure.
 

 Integration within a Web page 

1. Open the Web page in a text editor.

2. Copy and paste the JavaScript code below to the <head> section of the page.

See the JavaScript code

<script type="text/javascript" src="https://www.myselfhelpdomain.com/va-themes/EVSelfHelp.widget.min.js"></script>
<script type="text/javascript">
  var ezEvie = new EVSelfHelp.Widget(
   {
    "assistant-id": "MyAgentID",
    "lang": "en_US",
    "is-test": false,
    "config-url": "https://www.myselfhelpdomain.com/va-themes/ezv-panel/config.json",
    "quicklinks": ["Text1","Text2"],
    "contextvars":
     {
      "shoptions": "public",
      "theme": "evsh"
     },
    "custom":
     {
    "ff": "verdana",
    "fs": "14px",
    "f_color": "#1e1f1f",
    "title": "https://www.domain.com/logo.png",
    "subtitle": "Your virtual agent",
    "subtitle_color": "#ffffff",,
    "bg_color": "#6eb0da"
    "cb_name": "EVie",
    "cb_color": "#6eb0da",
    "cb_color_alt": "#2c658a",
    "user_color": "#ffffff", //f6f7f9
    "user_bgcolor": "#fa7268", //ffbb00
    "footer_color": "#ffffff",
    "placeholder": "Enter your question",
    "textarea_border": "1px solid #e8e8e8"
     }
   });
</script>

3. Specify the parameters in the JavaScript code that you inserted in the page.

  • www.myselfhelpdomain.com: Address of the domain hosting the Self Help project.
  • assistant-id: ID of the virtual agent.
  • lang: Language to be used.
        example  fr_FR, en_US
  • is-test: Indicates whether or not the agent should use the version of projects currently being edited. Note: If the value is true, then users must have the appropriate rights. Otherwise, the value  is false.
  • quicklinks: Predictive text suggested to users.
  • contextvars.shoptions: Public account when data must be made accessible to non-connected users.
  • contextvars.theme: evSH skin to be used for displaying procedures.
  • custom: Parameters for customizing the conversation panel.
Tags:
Last modified by Christine Daussac on 2020/05/06 20:22
Created by Administrator XWiki on 2019/12/10 17:17

Shortcuts

Recent changes
•  Service Manager
•  Service Apps
•  Self Help

Glossary

Powered by XWiki ©, EasyVista 2020