The Jira Service Management virtual agent
Step 1: Set up your Slack connection
The Jira Service Management virtual agent is currently available in Slack only. Microsoft Teams is coming soon.
The virtual agent lives in Slack, allowing you to meet your help-seekers where they already work to deliver always-on conversational support. If you’re new to providing support through Slack, check out our chat feature guide to learn more about capturing, managing, and resolving requests directly from your conversations.
Install Atlassian Assist
If you aren’t currently using Atlassian Assist for the Jira Service Management chat feature, you'll need to install the app before setting up the virtual agent. To do this, navigate to Project settings, then select Virtual agent. Click Add to Slack and follow the prompts to add the app to your Slack workspace.
Select a default request type
When the virtual agent is unable to resolve a request on your behalf, it will automatically create issues in Jira Service Management using a pre-selected default request type (don’t worry, you’ll have the option to override that with a specific request type later on if needed). To set your default request type:
- Select Settings from the virtual agent menu, then Basic settings at the top of the page.
- Under Virtual agent default request type, select the request type you want the virtual agent to use when creating issues. Please note, your default request type cannot have any required fields.
Set up Slack channels
To use the virtual agent, you’ll need to set up several Slack channels:
- Agent channel
- Request channel(s)
- Test channel
The agent channel is only accessible by agents and admins, and allows them to assign, triage, comment, and work on issues.
To create your agent channel:
- From Basic settings in virtual agent, select Create agent channel.
- Create a new private channel in Slack, or go to an existing one. (Hint: If the channel you want to use is public, convert it to private.)
- Add Assist to the channel using the /add command.
- When prompted, choose Agent channel and select your project.
- Use /invite to add agents and admins to the channel.
Request channels are where your customers can engage with the virtual agent to get help and create issues.
To add a request channel:
- In Slack, create a new channel, or go to the channel you want to use as a request channel.
- Add Assist to the channel using the /add command.
- When prompted, choose Request channel and select your project. This channel should now appear in your virtual agent settings in Jira Service Management under the Request channel tab.
The test channel allows you to see how your virtual agent works before turning it on for customers.
- From the navigation on the left, select Project settings, then Virtual agent, and then Settings.
- Select Settings at the top of the page.
- Under Test channel, select Create test channel.
Step 2: Configure your virtual agent
There are two primary ways to configure the Jira Service Management virtual agent, depending on the type and complexity of requests you’re looking to automate: pre-defined conversation flows and Atlassian Intelligence answers (AI answers). You can use one or both of these to help deflect tickets and deliver fast support to your customers.
1. Predefined conversation flows
A designated conversation progression is dependent on help seeker responses to the virtual agent.
Great for questions that:
- Require guided work/troubleshooting
- Require information collection and triaging
- Require an automated action via web request
Examples: Software access requests, reporting an incident, new hardware, procurement requests, onboarding workflows
2. Atlassian Intelligence answers
Uses generative artificial intelligence to search across your linked knowledge base spaces and answer your customer questions.
Great for questions that:
- Can be resolved by providing information or instructions
- Are covered in (or can be easily added to) your existing knowledge base articles
- Don’t usually need to be escalated to a human agent
Examples: Company holidays, WiFi access, BYOD setup, HR policies, password reset instructions
When a customer sends a message, the virtual agent will always first try to match it to an intent. If no intents match – either because you’ve chosen not to create any, or none of your existing ones are appropriate – it’ll then try to answer the question using Atlassian Intelligence answers. Below we’ll walk through these in more detail and show you how to set them up.
Set up pre-defined conversation flows
Pre-defined conversation flows are built based on the intents you create. Intents in the Jira Service Management virtual agent are how you represent a specific problem, question, or request that your virtual agent can help resolve for your customers. These are comprised of three key elements:
- Basic settings
- Training phrases
- Flows
This includes:
- Display name (this is how it might appear to customers when offered as one of multiple options)
- Internal description
- Confirmation question
Training phrases help your virtual agent match customer messages to this intent. You need a minimum of three per intent, but for best results, enter a wide variety of phrases used by your customers when asking for help with the subject of this intent
When the virtual agent detects an intent in a customer’s message, it asks them to confirm that the intent detected is correct. Once confirmed, the virtual agent starts the designated conversation flow for that intent.
Example intent
Let’s say a large number of your customers need help resetting their passwords for various software applications every day. You have instructions documented for each software program, but they’re hard to find, so customers frequently come to your chat platform to ask your support agents for help. This means that your agents spend a lot of their time repeatedly sending the same links to your customers.
In this scenario, you could create an intent with the display name Password reset, and train your virtual agent to recognize when customers are asking for help with resetting their password. Then you could configure the Password reset conversation flow so that the virtual agent:
- asks the customer which software program they want to reset their password for
- sends the customer a link to the relevant instructions
- asks whether the instructions helped them, and if not, creates an issue in Jira Service Management for a human agent to work on
How to create an intent
Intents can be created manually if your team already has an idea of the common issues you’d like to automate, but if not, the virtual agent makes it easy to discover and build intents with templates. Intent templates give you a list of training phrases and some basic settings to start with, and can be edited at any time. There are two types you can choose from:
- From your data: Powered by Atlassian Intelligence, templates are automatically created using recent issue data from your project. These are recommended based on expected coverage and are trained using real language from your customers. (Note: must activate Atlassian Intelligence to use this feature)
- Regular template: Templates are created based on common intents and issue categories used by teams like yours. (E.g. common HR questions, common IT requests, etc.)
To get started building an intent:
- From the navigation on the left, select Project settings, then Virtual agent.
- Select Intents, and then Create intent.
- To use a template, select Browse templates. Hover over a template to see its details. Select a template to use it.
- Enter (or edit) the Name, Description, or Display name.
- Enter (or edit) three Training phrases.
- Select Create.
Pro tip: 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 issues, 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 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 agent’s performance, you may want to come back to existing intents and edit your training phrases to make your virtual agent’s intent recognition even more effective over time.
Build a conversation flow
An intent flow behaves a lot like a normal conversation between two people: The virtual agent says something, waits for the customer to respond, then replies accordingly. A well-designed flow anticipates the things your customers might say and makes sure they get the help they need.
Flows are made up of a series of steps and branches, which visually represent each individual part of a conversation flow and the various paths they can take depending on how the customer responds to the virtual agent. Each branch can serve a different purpose or move the customer to a specific resolution.
There are five step options to choose from when building a flow:
- Offer choices: Offer the customer up to 10 pre-set options to choose from.
- Send message: Sends a message from the virtual agent to the customer – these messages can include some formatting and hyperlinks.
- Ask for information: Use this step when you’d like to collect information from customers to use as placeholders in a web request step, or to fill in a Jira field if an issue is created later in the conversation.
- Change request type/fields: Override the virtual agent default request type for any issues created after a specific point in a conversation flow.
- Web request: Use this step when you want to send an HTTP request to a URL, and choose a conversation branch based on the response. (Hint: This is how you can automate actions like granting software access.)
Pro Tip: Your virtual agent comes with standard flows out-of-the-box
Standard flows are reusable conversation flows that the virtual agent uses to complete basic, repetitive tasks, like greeting the customer, creating issues, or asking for clarification if something’s unclear. There are five standard flows that come out-of-the-box with your virtual agent:
- Greet: Posts a short, customizable greeting in response to a customer’s first message.
- Match intent: Quickly gets the customer into the correct intent’s conversation flow by asking a confirmation question, asking them to rephrase the question, presenting multiple intent options, or suggesting an answer from a knowledge base article.
- Resolve with CSAT: Ends a conversation when a customer indicates that their request has been resolved by the virtual agent by asking for a customer satisfaction (CSAT) score.
- Auto-close: Triggers when the customer is matched with an intent or AI answer, but there's been no activity in the thread for five minutes; prompts the customer to ensure they got the help they needed.
- Escalate: This lets the customer know that a request has been raised on their behalf and automatically creates an issue in Jira Service Management using the virtual agent default request type (unless it’s overridden during the conversation flow).
To create a new conversation flow:
- Select the intent you want to edit or create a conversation flow for and select Flow.
- Add, edit, or delete steps to build your conversation flow:
- To add a new step, select Add () in the place you want the step to appear, select the type of step you want to use, and then edit its details in the panel that appears on the right. Learn more about step types, or how to write effective messages for the virtual agent.
- To edit a step, select the step you want to edit, and then edit it in the panel that appears on the right.
- To delete a step, select the step you want to delete, then select Delete step at the bottom of the panel that appears on the right. If you delete a step that breaks off into multiple branches, only the branch on the left-hand side is retained.
Pro tip: Writing effective messages for your virtual agent
Well-written messages in your conversation flow can make a big difference in how your customers feel about their experience using the virtual agent. Here are a few ways to ensure the best help-seeker experience:
- Keep messages short. Your customers might be in a hurry, so adding too many pleasantries or unnecessary details could add frustration. Only include what they absolutely need to know.
- Be considerate, friendly, and neutral. While it’s tempting to have fun with the virtual agent, your customers may be coming to your channel for help with something urgent, sensitive, or potentially distressing. They also might be angry or frustrated if something isn’t working. A virtual agent that talks to them as if everything is perfect might make their mood or situation worse.
- Offer clear, helpful choices: This may seem obvious, but when adding an offer choices step, try to account for as many possible options customers might need. In many cases, it’s helpful to also include a choice that says “something else” (followed by more clarifying questions), or even raise a request (immediately followed by the escalate standard flow). These two additional choices give your customers a way to indicate they need something that isn’t covered in the choices you’ve offered or that they want to bypass the virtual agent and get help from one of your agents instead.
Once you’ve created your intent and finished building its conversation flow, it’ll automatically be active in your test channel. We recommend going to your test channel and testing your intent before activating it for your customers in a request channel.
Set up Atlassian Intelligence answers
If you’ve already got a knowledge base, setting up Atlassian Intelligence answers is a breeze – there are just a few quick steps you’ll need to take before you use it.
Connect your self-service knowledge base
First, you’ll need to ensure that you have a knowledge base linked to your project, either 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.
Pro tip: Double check your knowledge base permissions settings
Your linked knowledge base space needs to be set to All logged-in users under Who can view.
Activate Atlassian Intelligence for Jira Service Management
To take advantage of advanced AI features like Atlassian Intelligence answers, your org admin will need to turn on Atlassian Intelligence for each Atlassian product at admin.atlassian.com. All they have to do is choose which products they’d like to activate and accept the terms and conditions.
Activate Atlassian Intelligence answers in the virtual agent
Once Atlassian Intelligence is turned on for Jira Service Management and your knowledge base is ready to go, it’s time to activate AI answers in the virtual agent settings:
- From the navigation on the left, select Project settings, then Virtual agent.
- Select Settings, and then Basic settings if not already selected.
- Turn the toggle on next to Atlassian Intelligence answers, and then select Activate.
-
Navigate to Request channels in Settings
-
Turn the toggle on under AI answers next to the request channel you want to activate it for, and then select Activate.
Step 3: Test & launch your virtual agent
Once you’ve configured your virtual agent and you’re ready to launch it out into the wild, we recommend doing some simple testing and setting some goals you can measure to understand the performance and value of what you’ve built.
Take your virtual agent for a test drive
You can test your virtual agent using the test channel you created earlier – any intent flows you built will be live in there by default. Be sure to use feedback from testing to refine your conversation flows or update your training phrases, then test again one more time before activating the intent for customers.
Pro tip: Best practices for testing conversation flows
- Get as many people as you can to help you test and give feedback.
- Try and use every available branch.
- Get everyone to imagine they’re in different states of mind: in a rush, stressed about an upcoming presentation, or have difficult personal circumstances.
- Test your “edge cases” (unusual or highly specific problems and requests) and see what happens.
- Ask for help in as many different ways as you can to see if you need to update your training phrases.
Activate your virtual agent
To turn the virtual agent on or off in a request channel:
- From the navigation on the left, select Project settings, then Virtual agent, and then Settings.
- Select Request channels at the top of the page.
- Use the toggle next to the channel in which you want to turn the virtual agent on or off.
Pro Tip: You can activate or deactivate specific intents
Intents can have one of two statuses:
- Test – the intent is active in the test channel, but not in the customer-facing request channel.
- Live – the intent is active in both the customer-facing request channel and in the test channel.
When you create a new intent, its status is set to Test by default. When you’re ready to activate it for customers, you can change the status to Live.
When you turn the virtual agent on in a request channel, all intents set to a live status immediately begin working in that channel. Learn how to activate or deactivate specific intents in your request channels.
Measure
Once your virtual agent is live, it’s important to regularly check in on its performance to optimize your intent flows and AI answers. There are three main metrics the virtual agent tracks directly, which can be seen in the Intents panel of your virtual agent settings:
1. Resolution rate
The percentage of virtual agent conversations that were resolved without any human interaction.
2. Matched rate
The percentage of conversations that were matched to any intent. These could be resolved or handed off to an agent for further support.
3. Customer satisfaction (CSAT)
A rating out of five about a customer’s interaction with the virtual agent.
Want to give the virtual agent a try?
Tips and tricks
Form design