Set up the virtual service agent
20 min
By the end of this lesson, you'll be able to:
- Use AI answers in a virtual service agent
- Describe ways you can configure intents
- Set up virtual service agent to recognize questions
Automate support interactions using virtual service agent
Jira Service Management’s virtual service agent can help resolve common problems, answer frequently asked questions, and triage requests—potentially saving your agents hours of work.
Powered by Atlassian Intelligence, the virtual service agent automates support interactions from any of your request intake channels, such as the Jira Service Management portal, Microsoft Teams, or Slack, that you are using in Jira Service Management. This helps free up agent time and helps them deliver exceptional service at scale.
Space admins set up the virtual service agents.
In the virtual service agent, there are two core ways to automate responses to requests: Intents or Atlassian Intelligence answers, or AI answers. You can use one or both of these to automate responses to requests with the virtual service agent.
An intent is a specific problem, question, or request your virtual service agent can help resolve for your customers.
AI answers is a feature that uses generative AI to answer customer questions by searching linked knowledge base spaces.
Resolve requests using the virtual service agent
👇 Click the boxes to understand each method of automating the resolution of requests using a virtual service agent.
When a customer sends a message, the virtual service agent will always first try to match it to any existing intents. If no intents match, it’ll then try to answer the question using Atlassian Intelligence answers.
👇 This is how virtual service agents use intents.
You need to activate Atlassian Intelligence to use the virtual service agent.
Configure intents
When the virtual service agent detects an intent in a customer’s message, it asks them to confirm that the intent detected is correct. Once confirmed, the virtual service agent starts the conversation flow for that intent.
👇 Here's an example of a configured intent flow for setting up VPN.
Every conversation flow branch needs to end in one of two ways:
- Escalate standard flow, which creates a work item in Jira Service Management for a human agent to work on.
- Resolve standard flow, which marks the conversation as resolved and asks the customer for a satisfaction rating.
You need to be a space admin to end a conversation branch.
For guidance on creating or editing a conversation flow visit Jira Service Management support.
Admins can create intents manually for any work item or use intent templates, which include common IT requests or AI-powered templates based on historical work item data from the associated space.
You can choose from two types of templates:
- From your data: The virtual service agent leverages generative AI to recommend relevant intents based on your team’s historical ticket data, automatically populating essential settings such as descriptions and training phrases. Templates are automatically created using recent work item data from your space. These are recommended based on expected coverage and are trained using real language from your customers.
- Regular template: Default templates are created based on common intents and work item categories used by teams like yours, such as common HR questions and IT requests.
👇 Click the boxes to understand the two statuses of an intent.
Visit Jira Service Management support for help on creating, editing or deleting an intent.
Train your virtual service agent to recognize questions
When a customer's message matches an intent, the virtual service agent can initiate a conversation flow for assistance. However, it must first be trained to recognize intents in messages.
The agent develops a machine learning model based on a training data set tailored to your organization.
To train it, you must provide training phrases. The agent analyzes these phrases and uses machine learning to recognize the intent in actual customer messages, even if they don't exactly match your training phrases.
The training data set includes intents and training phrases:
- Created by your space admin
- Generated using your space's historical data that your space admin uses to create an intent
- Suggested by Atlassian for space admins to create an intent
- Created or modified by your space admin using any of the above methods
- The trained machine learning model for your organization is not shared with other Atlassian customers.
- Analysis of Jira Service Management work items and chat transcripts with your customers is used solely to enhance your experience.
How the virtual service agent’s data set is trained
To create an intent, start by defining it with a collection of common phrases that will be used to train the AI. Once the training is complete, you can switch the intent to live mode to handle incoming requests. As the AI processes these requests, it learns how customers typically phrase their questions. When a customer confirms with “Yes, it is the correct question,” the AI can start its analysis.
👉 For example: Many customers face daily VPN issues and need assistance. Although you have documented solutions, they are hard to find, leading customers to your chat platform for support, which consumes agents' time.
To improve this, create an intent titled “VPN issues” and train your virtual service agent to recognize requests for VPN help. You could also configure the conversation flow for these work items.
To establish an intent for VPN-related inquiries, start by including several common or similar training phrases that reflect this request type, such as:
- Broken VPN
- VPN issue
- VPN problem
Once the intent is activated, the AI will be able to recognize various phrases that convey similar meanings.
👉 For example: If a customer states, “My VPN is not connecting,” the AI, while still in the learning phase, can grasp the context and understand the customer's concern. It would then respond with, “Are you experiencing a VPN issue?” This capability allows the AI to identify relevant questions, even if the exact phrase was not part of the original training set.
Creating effective training phrases
- Use words and phrases most commonly used by your actual customers. Customers may ask for the same thing in very different ways — capture as many of these different ways of asking as you can! Read through past work items, and copy and paste the initial phrases your customers use when asking for help with the intent's main purpose.
- Add as many unique training phrases as you can. Avoid repetition, but add as many as you can to give the virtual service agent as much data as possible to train with. We recommend around 20 training phrases to start with, but you can add up to 100.
- Review and edit your training phrases as needed. As you watch the virtual service agent’s performance, you may want to come back to existing intents and edit your training phrases to make your virtual service agent’s intent recognition even more effective over time.
Configure AI answers
When AI answers is active, it first attempts to match an intent with the customer's initial message. If no high-confidence matches are found, it will answer the question using AI answers.
- Connect your self-service knowledge base: You can link a knowledge base to your space through Confluence or Jira Service Management’s native knowledge base. You can build out your knowledge base directly from Jira Service Management or integrate existing FAQs and docs you already have in Confluence.
👇 Here's an example of where you can link the knowledge base to your space.
2. Once your knowledge base is ready, activate AI answers in the virtual service agent settings:
a. In the sidebar, next to the space name, select More actions (represented by ···), then Space settings.
b. Select Channels & self service, then Virtual service agent.
c. Select AI answers.
c. Turn the toggle on for Atlassian intelligence answers.
👇 This is where the Atlassian Intelligence answers can be activated.
d. Select Activate in the pop-up.
👇 This is how the the confirmation dialog looks.
Set up chat in your Slack channel
- In the sidebar, next to the space name, select More actions (represented by ···), then Space settings.
- Select Channels & self service, then Virtual service agent.
- Select Settings.
- Select Manage under Slack to set up chat directly in Slack. When people ask for help in chat, your agents can create work items, respond, close requests, and more — all without leaving Slack.
👇 Here’s where you can set up chat for your Slack channel.
When creating knowledge base articles for virtual service agent, keep the following in mind:
- AI answers do not extract information from images.
- AI answers performs best with text not in tables in Confluence.
AI answer quality relies on the knowledge base's accuracy and clarity. A well-structured, up-to-date knowledge base enables better question deflection by the virtual service agent.